Add inline comments explaining data flow and memory ownership

Document the parsing pipeline (extract → tokenize → classify →
parse JSON → Phase 1 top-level prune → Phase 2 nested prune →
re-serialize → cleanup), memory ownership contract, why strtok_r
is mandatory under mod_h2, and the two-phase delete pattern for
json_object iteration safety.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>