fix(chess): parse compact PGN move numbers (1.e4) on import

[abdulwaarith0]

Summary

parse_moves() in backend/modules/chess/src/pgn.rs stripped move numbers with ^\d+\.+$, which only matches when the move number is its own whitespace-separated token (the spaced form 1. e4). A compact but valid PGN where the number is glued to the move (1.e4 e5 2.Nf3 Nc6, produced by many exporters/engines) tokenized to ["1.e4", "e5", "2.Nf3", "Nc6"]; the 1.e4/2.Nf3 tokens slipped through the filter and were rejected by the SAN parser.

This is live via the JWT-authenticated POST /games/import endpoint (backend/modules/api/src/games.rs:396), so importing a compact-format PGN returned a 422 "Illegal move" for a perfectly legal game.

Fix

Strip a leading move-number prefix from each token (^\d+\.+, no trailing $ anchor) instead of dropping whole-number-only tokens:

let move_number_prefix = Regex::new(r"^\d+\.+").unwrap();
...
.map(|token| move_number_prefix.replace(token, "").into_owned())
.filter(|token| !token.is_empty() && !result_regex.is_match(token))

This handles 1.e4, 1. e4, 1...e5, and 10.Ba2 uniformly.

Before / after

Input	Before	After
1.e4 e5 2.Nf3 Nc6 (compact)	["1.e4","e5","2.Nf3","Nc6"] → 422 ❌	["e4","e5","Nf3","Nc6"] → ✅
1. e4 e5 2. Nf3 Nc6 (spaced)	["e4","e5","Nf3","Nc6"] ✅	unchanged ✅

Tests

Added regression tests for compact, black-continuation (1...e5), and two-digit move numbers. All 9 pgn tests pass (6 existing + 3 new).

Closes #945