Semantic level cleanup

[benbrandt]

Extracts common logic into more defined areas:

• New Trim module to handle differing trimming logic
• Use unicode segmentation only as a fallback. This has the benefit of:
  • Being faster for most chunk sizes
  • Making sure we prefer the structure from the levels rather than unicode breakpoints (which causes different, but more correct output in Markdown splitting at small chunk sizes)

[codecov[bot]]

Codecov Report

Attention: Patch coverage is 99.46237% with 1 lines in your changes are missing coverage. Please review.

Project coverage is 99.73%. Comparing base (5b66e38) to head (50c4f14). Report is 1 commits behind head on main.

Files	Patch %	Lines
src/text.rs	0.00%	1 Missing :warning:

Additional details and impacted files

```diff @@ Coverage Diff @@ ## main #170 +/- ## ======================================= Coverage 99.72% 99.73% ======================================= Files 8 9 +1 Lines 1812 1863 +51 ======================================= + Hits 1807 1858 +51 Misses 5 5 ```

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.