Make REG_LITERAL match the empty pattern

[kevinushey]

These pull requests were generated with Claude Code, but were reviewed by me (@kevinushey) before posting. I'll respond personally to any feedback; please let me know if this is okay.

---

This makes REG_LITERAL match the empty pattern.

Problem. With REG_LITERAL, compiling an empty pattern succeeds, but the resulting regex never matches anything — including the empty string:

tre_regcomp(&re, "", REG_EXTENDED | REG_LITERAL);
tre_regexec(&re, "abc", 1, m, 0);   /* returns REG_NOMATCH */

Expected: an empty pattern matches the empty string at offset 0 in any subject, as it does without REG_LITERAL and in other regex engines.

Cause. The PARSE_ATOM empty-expression branch is explicitly skipped in literal mode (!(ctx->cflags & REG_LITERAL) && ...), so an empty pattern falls through without ever producing an atom.

Fix. In literal mode at end of pattern, emit an EMPTY literal node, mirroring the non-literal branch. The guard uses ctx->re >= ctx->re_end so it also covers tre_regncomp(&re, "", 0, ...).

Testing (Windows x64, mingw-w64 gcc 14):

compile "" LITERAL                           PASS
"" matches in "abc" at (0,0)                 PASS   (unpatched: FAIL)
"" matches in "" at (0,0)                    PASS   (unpatched: FAIL)
regncomp len=0 LITERAL matches               PASS   (unpatched: FAIL)
literal "a.c" still matches only "a.c"       PASS
non-literal "" still matches (baseline)      PASS

Provenance. A variant of this fix has been carried in GNU R's vendored copy of TRE since 2009, where REG_LITERAL backs R's fixed = TRUE matching.