du: use uutils-args to fix some bugs due to usage of clap

[BenWiederhake]

This is a proof-of-concept that using uutils-args would measurably improve compatibility with GNU coreutils, while still passing all existing tests.

Context:

• We want to switch away from clap eventually: New argument parser #4254
• We already avoid using clap in expr: expr: Get rid of clap for performance reasons #7352
• This heavily builds on the existing du cli-parser prototype: implement proof of concept for du -bkB uutils-args#114
• (du: block size parsing #7738 is unrelated)

However, this probably shouldn't be merged as-is:

• This PR uses a specific commit of uutils-args, we should probably consider doing releases of uutils-args at some point. Who can decide that? What are the relevant things to consider here?
• While uutils-args significantly improves a lot of things (i.e. handling early exit due to --help), there are some edge cases it handles wrong: Only sometimes, a late --help should override other errors uutils-args#130 We should resolve this issue first, as it might change the architecture.
• Before pushing uutils-args on other people, it should become a little nicer with its error messages first: Typo in #[arg()] leads to opaque error message uutils-args#129 I don't know enough about proc macros, can someone help here?
• At some point in time (now? later?), the signature fn uu_app() -> Command needs to change. See discord for that discussion; I guess there's going to be vivid discussion.

On the other hand, this is a valid approach, and could in theory be used as-is, hence I didn't mark this as "draft".

What do you think?

[github-actions]

GNU testsuite comparison:

GNU test failed: tests/du/inodes. tests/du/inodes is passing on 'main'. Maybe you have to rebase?
GNU test failed: tests/du/max-depth. tests/du/max-depth is passing on 'main'. Maybe you have to rebase?
GNU test failed: tests/du/threshold. tests/du/threshold is passing on 'main'. Maybe you have to rebase?
Skip an intermittent issue tests/misc/usage_vs_getopt (fails in this run but passes in the 'main' branch)
Skip an intermittent issue tests/tail/inotify-dir-recreate (fails in this run but passes in the 'main' branch)
Skipping an intermittent issue tests/timeout/timeout (passes in this run but fails in the 'main' branch)

[BenWiederhake]

Many changes since last push. Let's hope I remember them all:

• This now uses Make 'complete' available to procmacro's output by conversion to module uutils-args#135 , which is necessary because it won't compile otherwise.
• I now also include -h and --si in the block size parsing, because these three options overwrite each other, as demonstrated by the new test test_du_order_si_h_b.
• I moved value parsing to the general parsing stage, not afterwards. One effect is that -B banana -B 1 is no longer accepted, identical to GNU behavior.
• I also found several new edge cases we currently handle wrongly. Some of them are fixed just by switching to uutils-args, see test_du_error_precedence and test_du_blocksize_bytes_order
• New edge cases that I couldn't fix but needed to interact with are linked or marked with // FIXME.
• The generated errors are surprisingly bad. I probably did something wrong. The worst offenders are:
  • du: Invalid value '000' for '-B': (note the trailing space and missing continuation after the colon)
  • du: Invalid value 'banana' for '-B': 'banana' (note the pointless duplication)

In other words, this PR can't be merged as-is, but it's a nice demonstration to what uutils-args can do.

[github-actions]

GNU testsuite comparison:

GNU test failed: tests/du/inodes. tests/du/inodes is passing on 'main'. Maybe you have to rebase?
GNU test failed: tests/du/max-depth. tests/du/max-depth is passing on 'main'. Maybe you have to rebase?
GNU test failed: tests/du/threshold. tests/du/threshold is passing on 'main'. Maybe you have to rebase?
Skip an intermittent issue tests/misc/usage_vs_getopt (fails in this run but passes in the 'main' branch)