[cmd] Emit errors instead of hiding flags

[Szelethus]

One of the great annoyances with any tool, but in particular with CodeChecker is that we have a habit of hiding, suppressing stuff instead of just telling the user whats up. Such is the case with --stats, --ctu, etc. These flags can only be used if the clang CodeChecker found supports statistics generation or CTU analysis. However, if you try to invoke CodeChecker with an incompatible flag, you get something like this:

$ CodeChecker analyze compile_commands.json  --stats -o output
usage: CodeChecker [-h] {analyze} ...
CodeChecker: error: unrecognized arguments: --stats

That is nuts! Not only do we emit this message, it doesn't even show up in --help. How is a user supposed to know about z3 analyses, if we never show it in the first place? If we did, a user might get a hold of a clang that supports it, there is literally zero reason to just outright hide it.

With this patch, all flags that depend on an analyzer tool having a specific capability are displayed by default, and a helpful error message is shown if its not available. Here is an example:

$ CodeChecker analyze compile_commands.json  --stats -o output
[ERROR 2025-02-20 15:04] - Statistics options can only be enabled if clang has statistics checkers available!
[INFO 2025-02-20 15:04] - CodeChecker has not found Clang Static Analyzer checkers statisticsCollector.ReturnValueCheck and statisticsCollector.SpecialReturnValue
[INFO 2025-02-20 15:04] - hint: Maybe CodeChecker found the wrong analyzer binary?

The change is as trivial as it looks, the only thing I changed is that some statistics related option's default value now depends on whether clang is stats capable (similarly to how its always been done for ctu flags), and since the z3 tests checked z3 compliance from CodeChecker's own error message, that needed adjustment as well. In the Z3 tests, the command that runs CodeChecker anayze were moved below the part where the compile_commands.json is copied.

[Szelethus]

As a sidenote: I tried creating a custom argparse action that prematurely exits as the argument is parsed, but that package is so hard to work with I figured it takes more effot than its worth. In particular, I couldn't revert to "store_true" action type after my own checks were done, I couldn't even iterate through the subarguments of an argument group, etc. Hence the not-so-clever patch.

[bruntib]

I like this improvement!

[bruntib]

"CodeChecker check" is using "CodeChecker analyze" by calling its main() function like a library code. For this reason there is no need to duplicate this logic.

[bruntib]

Sorry, I wanted to request for some changes :(
Also, please check the failing tests, too.