Bound a w or W blank run at the end of the string buffer

[LouisDeconinck]

libmagic behavior

A blank in a w or W value skips a whitespace run in the candidate bytes, but the run is not unbounded: mcopy copies the candidate into a MAXstring-byte union (file/src/file.h:179, file/src/softmagic.c:1478-1511), mconvert forces that buffer's last byte to NUL before the comparison (file/src/softmagic.c:1238), and file_strncmp stops the run at the end of the buffer via eb = b + maxlen (file/src/softmagic.c:2070-2072, 2102-2121). A run long enough to reach the buffer's last byte therefore ends there, and the next byte of the value compares against the terminating NUL and fails.

Divergence

PolyFile compiled each blank to an unbounded [ \t\n\v\f\r]+ (or *), so a string/W or string/w test matched a run of any length:

$ printf '0\tstring/W\tA\\ B\tfound\n' > w.magic
$ python3 -c "open('long.bin','wb').write(b'A' + b' ' * 126 + b'B')"
$ LC_ALL=C TZ=UTC file -b -m w.magic long.bin
ASCII text, with no line terminators

PolyFile reported found for that file and for a run of any length.

Fix

StringMatch.matches now runs the compiled pattern over a reproduction of libmagic's buffer: at most MAX_STRING_BYTES - 1 bytes of the candidate plus the terminating NUL (data[:limit].ljust(MAX_STRING_BYTES, b"\x00")). A str_range (string/N) shortens the copy the same way mcopy shortens it (softmagic.c:1480-1481). search is untouched: libmagic reads that buffer in place and bounds the run at the end of the remaining buffer instead (file/src/softmagic.c:2363-2364), which the search path already does.

Test

test_a_blank_run_stops_at_the_end_of_the_string_buffer pins the boundary for both spellings: 125 blanks match, 126 or more fall back to the text result, and a padded file shows the bound is the buffer rather than the file's length. A search/200/W case pins that a search still matches a longer run.

Verified against file 5.46 and the bundled 5.48 source: both flags, negated tests (string/W !A\ B reports found on a 126-blank file, as libmagic does), and multi-blank values agree at every boundary probed.

pytest tests/test_magic.py: 220 passed. flake8 on the touched files reports only the warnings already present on master.

Fixes #3571