feat(async): complete AsyncMlb endpoint coverage (stats, persons, schedule, gamepace)

[Mattsface]

Closes #305.

Ports the last seven Mlb methods to AsyncMlb. Both clients now expose 41
public methods; the only remaining difference is that close() is spelled
aclose().

What's here

Two commits.

7d2f0ec — the stats group. get_stats, get_player_stats,
get_team_stats, get_players_stats_for_game.

113a8c0 — the last three. get_persons, get_scheduled_games_by_date,
get_gamepace.

Scope was deliberately held to strict parity. No new types, constants, or
validation — the usability problems these endpoints have are real but are
tracked separately in #326 rather than mixed into a port.

The bug worth reviewing carefully

Mlb.get_gamepace builds its request like this:

self._mlb_adapter_v1.get(endpoint=f'gamePace?season={season}', ep_params=params)

The season rides in the endpoint string; sportId and everything else arrive
via ep_params. That works on the sync side because Requests merges an
endpoint's existing query with the params it's given. HTTPX replaces it:

requests: GET /api/v1/gamePace?season=2021&sportId=1
httpx:    GET /api/v1/gamePace?sportId=1

Copying the sync idiom into AsyncMlb would therefore have dropped the season
silently — no exception, just game pace data for the wrong scope. AsyncMlb
passes the season as an ordinary param instead, producing an identical request.

Two follow-on notes:

• get_awards is the only other method that embeds a query in its endpoint
  string (awards/{id}/recipients?). That query is empty, so merge-vs-replace
  makes no difference there and it was never affected.
• Mlb is left alone. Its idiom is fragile but correct under Requests, and
  changing it is out of scope for a parity port.

The test helper had the same blind spot

tests/test_async_mlb.py::assert_matches_sync derives each expected request
from Mlb itself, which is what makes these ports trustworthy. But it compared
url.path against the raw endpoint string, so gamePace?season=2021 was never
parsed as a query and the drift above would have passed review.

It now splits an endpoint's embedded query and folds it into the expected
params — the same thing Requests does — so the expectation is the merged query
either client has to end up sending, regardless of how it built the URL. That
also subsumes the get_awards trailing-? special case it had been carrying.

Other changes

Three new shared parsers, following the pattern the rest of the async port
uses: _parsers/stats.py::parse_split_stats,
_parsers/schedules.py::parse_scheduled_games, and _parsers/gamepace.py.
Each replaces inline logic that existed only in mlb_api.py; the stats one
alone collapses four copy-pasted blocks into one.

A sync bugfix. Mlb.get_players_stats_for_game accepted **params and
never passed ep_params, so every caller-supplied keyword was silently
discarded. Both clients now forward them. This changes sync behavior, but in
the direction its own signature already advertised — agreed on before
implementing.

Two preserved quirks, both documented rather than corrected:

• get_scheduled_games_by_date is annotated list[ScheduleGames] but returns
  None when nothing selects a date. Asserted explicitly in the parity suite.
• The stats methods return {} for a 404, for a body with no stats, and for
  an unrecognized stat type or group. That last case is the subject of Make the stats endpoints easier to use #326.

Docstring corrections on Mlb.get_players_stats_for_game, which described
person_id as "the team id", game_id as "list of stat types", and whose
example called get_player_stats_for_game — not a method.

Testing

poetry run pytest tests/ --ignore=tests/external_tests

• Before: 974 passed
• After: 1057 passed (+83)

poetry run pytest tests/external_tests/

148 passed, 1 skipped against the live API — this is what covers the sync
refactors, since create_split_data and the schedule/gamepace parsers now run
through new code paths on real payloads.

Also verified directly against the live API that get_gamepace, get_persons,
and get_scheduled_games_by_date return objects equal to their sync
counterparts, rather than trusting the canned fixtures alone.

New coverage: 8 stats-parser tests, 7 gamepace-parser tests, 4
scheduled-games-parser tests, 24 async client tests, and 33 parity tests.

Risk

Low for the async additions — new methods, nothing existing changes shape.

Moderate for the two sync-side changes: the get_players_stats_for_game param
fix is a real behavior change, and the four stats methods plus
get_scheduled_games_by_date/get_gamepace now route through extracted
parsers. The external suite is the check that matters there and it's green.

🤖 Generated with Claude Code