[HongTin2104]

Summary

Reusing a Markdown() instance was retaining entries in _escape_table and _code_table across documents. These tables grew with each call, causing two problems:

Wrong output: A backslash before a URL that appeared in an earlier document was silently dropped, because the URL hash was already in _escape_table from that earlier call.

Quadratic slowdown: _escape_table and _code_table accumulated every URL and code span ever seen. With 40 documents of 200 links and 200 code spans each the 40th document took ~37× longer than the 1st.

Root cause

reset() cleared urls, titles, html_blocks, html_spans, list_level, and extras, but never touched _escape_table or _code_table.

Fix

Extract the table initialisation from __init__ into _reset_tables() and call it from both __init__ and reset(). The call in reset() is placed before _setup_extras() because the middle-word-em extra reads _escape_table.

The convenience function markdown2.markdown() is unaffected (it creates a new instance per call).

Verification

import markdown2
md = markdown2.Markdown()
md.convert('[a link](https://example.com/page)')
print(md.convert(r'Type \https://example.com/page to see it.'))
# Before: <p>Type https://example.com/page to see it.</p>
# After:  <p>Type \https://example.com/page to see it.</p>  (matches a fresh instance)

Added test/test_reuse_instance.py with 3 tests: backslash escape correctness, code span isolation, and a performance guard.

Closes #763