stat: avoid printf formatter panics

[wondr-wclabs]

What changed

This updates stat --printf formatting so large but finite widths/precisions no longer go through Rust's dynamic-width formatting machinery.

The change covers the reported panic paths in #12594:

• field width padding such as %111111.1d and %111111.1f
• numeric precision zero-padding such as %.100000d, %.100000f, %.111111s, and %.111111a

Why

The panic was not caused by stat metadata itself. It came from formatting calls like {result:>width$} and {num:0>p$x}. Rust's formatter has an internal bound for dynamic width/precision and panics with Formatting argument out of range for values that are still finite and reasonable for command output tests.

Rather than lowering accepted stat --printf widths, this writes padding in chunks and builds numeric zero-padding explicitly. That keeps large finite output possible while avoiding formatter panics. For string widths, the helper counts chars() before padding so it preserves the old Rust formatter width behavior for multibyte UTF-8 strings, instead of accidentally switching to byte-width alignment.

Validation

• cargo test --no-default-features --features stat --test tests test_stat::test_printf_large_width_and_precision_do_not_panic -- --nocapture
• cargo test --no-default-features --features stat --test tests test_stat -- --nocapture
• cargo test -p uu_stat -- --nocapture
• cargo clippy -p uu_stat -- -D warnings
• cargo clippy --no-default-features --features stat --test tests -- -D warnings
• cargo fmt --check
• git diff --check

Fixes #12594.

[github-actions]

GNU testsuite comparison:

Note: The gnu test tests/cut/bounded-memory is now being skipped but was previously passing.