fix(cli): exit 1 when check cannot read its input

[dekobon]

Fixes #1060.

The reported bug

validate_and_resolve_file bumped files_dispatched before
attempting the read, so the "no input files matched" guard in
run_check saw a non-zero count on a run that read nothing:

$ chmod 000 violator.c
$ bca check --no-config -p . --threshold cyclomatic=2
error processing ./violator.c: Permission denied (os error 13)
$ echo $?
0

The counter now moves only for files that were actually read, and read
failures are tallied separately in Config::read_failures. Any
unreadable input exits 1 (tool error), not 2 (gate breach) — a gate
that analysed less than its input has no verdict to report.

The half that wasn't reported

Reordering the counter alone would have left the bug live for files of
three bytes or fewer. read_file_with_eol decided "too small to
parse" from a bare fs::metadata and returned Ok(None) before
File::open — and stat succeeds on a file the process cannot read, so
a tiny unreadable file was counted as empty:

$ printf 'a;\n' > small.c && chmod 000 small.c
$ bca check --no-config -p . --threshold cyclomatic=1 ; echo $?
0

Note this produced no diagnostic at all — not even the
error processing … line from the original report — so it was strictly
worse than the reported symptom, and it contradicted the function's own
documented contract ("Returns any std::io::Error surfaced by
File::open … lacks read permission"). The shortcut now confirms
readability by opening the file first; the open is skipped for anything
that is not a regular file, so the function still never blocks on a
FIFO. Behaviour is unchanged for readable files of any size.

Scope notes

• bca init shares the guard. scaffold_baseline produces its
  baseline by driving run_check, so an unreadable file aborts init
  rather than pinning a baseline that silently under-records the debt in
  the file it could not read. Both guards' messages are now prefixed
  bca: instead of bca check:, which misattributed an init failure
  to a subcommand the user never ran.
• Not suppressed by --no-fail, which suppresses threshold
  failures, not broken input — consistent with the pre-existing
  empty-input guard. Both guards run before the gate is evaluated and
  before --write-baseline, so a partial run never records a baseline.
• This can turn a currently-green CI red. A repo with any unreadable
  regular file under --paths (a stale docker-built target/, a
  node_modules owned by another uid) now fails until the path is
  excluded. That is the intended contract, and it matches what
  commands/README.md and man/bca.1 already documented ("unreadable
  input" → 1).
• Other subcommands are unchanged — bca metrics on an unreadable
  file still exits 0. Tracked in fix(cli): read failures exit 0 for every subcommand except check #1098, where bca diff --since is the
  sharper case (a dropped file reads as an added/removed function rather
  than an I/O error).

Tests

Every test was verified by perturbing the production line and
confirming it was the only failure in its suite:

• dispatch.rs — an unreadable file bumps read_failures and not
  files_dispatched; a readable one bumps the opposite; an empty
  file still counts as dispatched, so the reorder did not narrow the
  documented semantics.
• check_thresholds.rs — all-unreadable, partially-unreadable (two
  files, pinning the plural summary), tiny-unreadable, and
  --no-fail-does-not-mask.
• init.rs — init refuses to baseline a partially-readable tree.
• tools_tests.rs — a tiny readable file is still Ok(None); a tiny
  unreadable one returns PermissionDenied.

The unix-only tests each probe the real capability (fs::read on a
mode-000 file) rather than the uid, so a privileged runner skips instead
of failing, and #[cfg(unix)] sits on the fn rather than an inner
block (lesson #40).

make pre-commit is green. .bca-baseline.toml was refreshed in the
same change for the one metric that moved
(read_file_with_eol.nexits 6 → 7).

[codecov]

Codecov Report

❌ Patch coverage is 80.00000% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 98.08%. Comparing base (9f1d99d) to head (67f0eb7).

Files with missing lines	Patch %	Lines
src/tools.rs	80.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1099      +/-   ##
==========================================
- Coverage   98.08%   98.08%   -0.01%     
==========================================
  Files         272      272              
  Lines       67791    67795       +4     
  Branches    67361    67365       +4     
==========================================
+ Hits        66493    66496       +3     
- Misses        887      888       +1     
  Partials      411      411              

Flag	Coverage Δ
rust	98.07% <80.00%> (-0.01%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
src/tools.rs	93.66% <80.00%> (-0.26%)	⬇️

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.