Improve tidy-requires

Motivated by jackfirth/resyntax#432, as well
as inspired by racket-review's use of syntax classes.

Define syntax classes closely following the require grammar. Use them
to drive the matching and simplification, producing attribute values
that are `sortable` structs, from which we reconstruct the syntax.

Because the macro-debugger-lib drop analysis works in terms of <module
phase> tuples, the root-module-path syntax class is the locus -- it
may produce #f instead of a `sortable` to indicate elision. This can
percolate up through other specs, since they all build directly or
transitively on root-module-path.

Simplify gratuitous specs either reducing to the simpler equivalent
spec or eliding entirely. e.g. (only-in m) => m or (combine-in) => #f.

No longer "explode" multi-in forms when denormalizing and attempt to
recover them when normalizing. Instead: Preserve multi-in forms,
modifying only for drops. In simple cases like (multi-in a b (c d)),
we can modify the multi-in form itself. In the general case -- any
Cartesian product -- we must wrap in a subtract-in spec.

Change require tidying to work on full file syntax; requires are
tidied in all modules. The racket-{trim base}-requires add/drop
required modules only in the file's root module, because the
macro-debugger lib's analysis only works for that module. As a result
of this change, the front end no longer needs to gather and supply a
list of require forms. Instead it always just supplies a file path.
The back end commands return a list of textual changes -- inserts,
deletes, and replaces -- for the front end to make to the buffer. This
seems like a cleaner and more appropriate division of labor.

Improve some doc strings in substance, and prefer active to passive
voice.

doc/racket-mode.texi

@@ -1069,37 +1069,24 @@ Unfold (show) all test submodules.
 
 @kbd{M-x}  @code{racket-tidy-requires}
 
-Make a single ``require'' form, modules sorted, one per line.
+Make a single, sorted ``require'' form.
 
-The scope of this command is the innermost module around point,
-including the outermost module for a file using a ``#lang'' line.
-All require forms within that module are combined into a single
-form. Within that form:
+The scope of this command is the innermost module around point --
+whether an explicit submodule form or the outermost module for a
+file that has a ``#lang'' line.
 
-@itemize
-@item
-A single subform is used for each phase level, sorted in this
+Merge all require forms within that module to one form.
+
+Use a single require-spec for each phase-level, sorted in this
 order: for-syntax, for-template, for-label, for-meta, and
 plain (phase 0).
 
-@itemize
-@item
-Within each level subform, the modules are sorted:
-
-@itemize
-@item
-Collection path modules -- sorted alphabetically.
-
-@item
-Subforms such as only-in.
+Within each phase-level, sort require-specs by module name.
 
-@item
-Quoted relative requires -- sorted alphabetically.
-@end itemize
-@end itemize
-@end itemize
+Format at most one module per line.
 
-At most one required module is listed per line.
+Simplify gratuitous require-specs. For example reduce (only-in m)
+to m and elide (combine-in).
 
 See also: @ref{racket-trim-requires} and @ref{racket-base-requires}.
 
@@ -1108,16 +1095,28 @@ See also: @ref{racket-trim-requires} and @ref{racket-base-requires}.
 
 @kbd{M-x}  @code{racket-trim-requires}
 
-Like @ref{racket-tidy-requires} but also deletes unnecessary requires.
+Like @ref{racket-tidy-requires} but also delete unnecessary requires.
+
+Use macro-debugger/analysis/check-requires to analyze.
+
+The analysis:
+
+@itemize
+@item
+Needs the @code{macro-debugger-lib} package.
 
-Note: This only works when the source file can be fully expanded
-with no errors.
+@item
+Only works when the source file can be fully expanded with no
+@end itemize
+errors.
 
-Note: This only works for requires at the top level of a source
-file using #lang. It does NOT work for require forms inside
-module forms. Furthermore, it is not smart about module+ or
-module* forms -- it might delete top level requires that are
-actually needed by such submodules.
+@itemize
+@item
+Only works for requires at the top level of a source file using
+@end itemize
+#lang -- not for requires inside submodule forms. Furthermore, it
+is not smart about module+ or module* forms -- it might delete
+outer requires that are actually needed by such submodules.
 
 See also: @ref{racket-base-requires}.
 
@@ -1128,24 +1127,16 @@ See also: @ref{racket-base-requires}.
 
 Change from ``#lang racket'' to ``#lang racket/base''.
 
-Adds explicit requires for imports that are provided by
-``racket'' but not by ``racket/base''.
-
-This is a recommended optimization for Racket applications.
-Avoiding loading all of ``racket'' can reduce load time and
-memory footprint.
+Using ``racket/base'' is a recommended optimization for Racket
+applications. Loading all of ``racket'' is slower and uses more
+memory.
 
-Also, as does @ref{racket-trim-requires}, this removes unneeded
-modules and tidies everything into a single, sorted require form.
+Add explicit requires for imports that are provided by ``racket''
+but not by ``racket/base''.
 
-Note: This only works when the source file can be fully expanded
-with no errors.
-
-Note: This only works for requires at the top level of a source
-file using #lang. It does NOT work for require forms inside
-module forms. Furthermore, it is not smart about module+ or
-module* forms -- it might delete top level requires that are
-actually needed by such submodules.
+Also do the equivalent of @ref{racket-trim-requires} and
+@code{nil}. See those commands for additional notes
+and caveats.
 
 Note: Currently this only helps change ``#lang racket'' to
 ``#lang racket/base''. It does not help with other similar
@@ -1162,28 +1153,31 @@ Add a require for an identifier.
 Useful when you know the name of an export but don't remember
 from what module it is exported.
 
-At the prompt:
+1 At the prompt, you may:
 
+@itemize
+@item
 Use @kbd{<XF86Forward>} or @kbd{<next>} or @kbd{M-n}  to load the identifier at point.
-You may also need to @kbd{C-e} or @kbd{<end>}  to see candidates.
-
-Or type anything.
+@end itemize
+You might also need to @kbd{C-e} or @kbd{<end>}  to see candidates.
 
-After you choose:
+@itemize
+@item
+Or, type anything.
 
-The identifier you chose is inserted at point if not already
-there.
+@item
+After you choose, this command will:
 
-A ``require'' form is inserted, followed by doing a
-@ref{racket-tidy-requires}.
+@item
+Insert the identifier at point if not already there.
 
-When more than one module supplies an identifer with the same
-name, the first is used -- for example ``racket/base'' instead of
-``racket''.
+@item
+Insert a ``require'' form and do @ref{racket-tidy-requires}.
+@end itemize
 
-Caveat: This works in terms of identifiers that are documented.
-The mechanism is similar to that used for Racket's ``Search
-Manuals'' feature. Today there exists no system-wide database of
+Caveat: This works only for identifiers that are documented. The
+mechanism is similar to that used for Racket's ``Search Manuals''
+feature. Today there exists no system-wide database of
 identifiers that are exported but not documented.
 
 @node racket-indent-line