Range-check rc_dup_str length to prevent silent int truncation

[billdenney]

rc_dup_str (src/shared.c) computed the segment length as
int l = e ? e-s : (int)strlen(s);
where the source pointers come from the dparser parse tree. When the source span exceeds INT_MAX bytes, the implicit ptrdiff_t-to-int or size_t-to-int cast silently truncates the length to a wrong value (often negative). Downstream addLine(&_dupStrs, "%.*s", l, s) then either reads past the buffer or copies the wrong number of bytes.

Replace the silent cast with an explicit range check on both branches (ptrdiff_t > INT_MAX and size_t > INT_MAX) and raise a clean R error when out of range. Also add a thread-safety comment documenting that the file-scope parser globals are intentionally not mutex-protected because R's interpreter is single-threaded.

Adds tests/testthat/test-mem-rc-dup-str.R as a regression test. The INT_MAX-input test is skip()ed because it requires ~2 GB of free RAM.

[codecov]

Codecov Report

❌ Patch coverage is 77.77778% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 89.44%. Comparing base (ca1ffea) to head (e68fe97).

Files with missing lines	Patch %	Lines
src/shared.c	77.77%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #34      +/-   ##
==========================================
- Coverage   89.44%   89.44%   -0.01%     
==========================================
  Files          60       60              
  Lines        6360     6368       +8     
==========================================
+ Hits         5689     5696       +7     
- Misses        671      672       +1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.