use normal entrypoints into base's display pipeline

[rokke-git]

Base and SparseArrays had some funky action-at-a-distance stuff going on, this fully disentangles the printing functions. this is what I really came over here for, the type thing was just me hoping to simplify things a bit. didn't end up effecting much, but does add the circular_reference line.

it also makes SparseArrays a little smarter about when to swap to braille output, and actually uses the whole screen instead of just the left half. it's also theoretically doing less work now, but it's just a printing function so whatever. the width of displayed zeros also doesn't depend on the type anymore, and is always centered.

since this removes the error that used to come up during printing, I added in a new warning to replace it.

fixes: #21, #22, #233, #618

[rokke-git]

lol, guess the doctests are a good way of displaying what's changed...

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 84.45%. Comparing base (751387e) to head (1eb9e69).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #736      +/-   ##
==========================================
+ Coverage   84.40%   84.45%   +0.05%     
==========================================
  Files          13       13              
  Lines        9374     9394      +20     
==========================================
+ Hits         7912     7934      +22     
+ Misses       1462     1460       -2     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[rokke-git]

since I'm already here, started looking through existing display issues.
now fixes: #21, #22, #233, #618

[rokke-git]

this last push just changed the Matrix(x) conversion call as specified in #618 from dest[isrc[row, col]] = val to dest[isrc[row, col]] += val, which technically adds a get to a potentially hot loop (ie, not display, actual computation).

if this should instead be actually benchmarked I can remove this change; or if we don't care it can remain fixed like this. the dest matrix was already initialized with basically zero(eltype(x)), so Matrix(x) would have already been failing for non-numeric types, no change there

[rokke-git]

I'd say this last commit fixes #420 as well; it's much more clear that it's an image of your actual matrix now that there's no aliasing

[rokke-git]

ok, I made a couple final tweaks and then sorted all the changes into 3 sensible commits. only have it separated out so it's easier to look at, can be squashed for merging; the first commit message counts for all three. any other changes you want me to make?