docs: add Coming-from-Python section + fix stale counts

[conorbronsdon]

Summary

Two documentation changes:

1. Coming from Python — a short side-by-side section (near the top, after
   the intro) mapping the familiar Python API to this library's real API.
   Every snippet was extracted from this repo's own examples/ and test/;
   nothing was invented.
2. Accuracy fixes — Suite copy Ten -> Eleven and added mojo-xml link. Test count re-verified by counting: 32 (matches README).

Test counts were re-verified by actually counting the test functions in the
repo (not trusting the prior text).

From an automated review (personal-context#62); drafted by Claude Code — please review the Coming-from-Python snippets for API accuracy.

[conorbronsdon]

🤖 Independent Claude review: Ready to merge.

Verified every Mojo API in the new "Coming from Python" table against src/diff/ — all real, no invented APIs:

• unified_diff(a, b, fromfile, tofile, n) → src/diff/diff.mojo:266 def unified_diff(a_text, b_text, from_file="a", to_file="b", context=3). The "takes whole strings and splits for you" note is accurate — it calls splitlines_keepends internally. Matches examples/diff_files.mojo.
• ratio(...) → src/diff/diff.mojo:179 def ratio(a: List[String], b: List[String]).
• get_opcodes(...) → src/diff/diff.mojo:152 def get_opcodes(a: List[String], b: List[String]).
• splitlines_keepends(s) → src/diff/diff.mojo:28 def splitlines_keepends(text: String) -> List[String] — return type feeds ratio/get_opcodes correctly.

All four are exported from src/diff/__init__.mojo. Python-side snippets (difflib.unified_diff, SequenceMatcher(None, a, b).ratio(), .get_opcodes()) are all real, idiomatic difflib.

Accuracy checks:

• "32 tests" — correct: 32 def test_* fns in test/test_diff.mojo.
• "Eleven libraries" — correct: 10 linked repos + mojo-diff itself = 11.
• New mojo-xml cross-link resolves; repo description ("ElementTree-shaped DOM") matches the README's characterization.

No blocking issues.

[conorbronsdon]

Docs review (sonnet): SHIP

CHANGELOG "32 tests" verified (32 def test_). Table matches src/diff/diff.mojo — unified_diff(...) takes whole strings (correctly contrasted with Python's list-based), ratio/get_opcodes take List[String]. Suite Ten→Eleven correct. Cosmetic only: the table labels a param n where the source names it context (positional, so harmless).