Fix #2462: allow MATCH Input when Output has no MATCH

[bpostlethwaite]

Fixes #2462.

Two years of comments on that issue all land in the same place: people want to use MATCH in an Input/State without repeating the same MATCH keys on the Output. This PR does exactly that.

Behavior change

You can now pair a MATCH input with a fixed-id Output, no Output, or a wildcard Output that only uses ALL:

@callback(
    Output("notifications", "children"),
    Input({"type": "grid", "id": MATCH}, "cellValueChanged"),
)
def on_edit(event):
    return f"edited: {ctx.triggered_id}"

Rules kept intact:

• All Outputs of a callback must still share the same MATCH keys.
• ALLSMALLER still requires a corresponding MATCH in an Output — it has no reference otherwise.

What's in the diff

1. Validation (dependencies.js::findMismatchedWildcards) — skip the Input/State MATCH-subset check when the Output has no MATCH keys.
2. Resolution (dependencies.js::addAllResolvedFromOutputs + dependencies_ts.ts::getCallbacksByInput) — when the Output lacks MATCH, the triggering Input's MATCH values become the callback's anyVals so simultaneous MATCH triggers each get a distinct resolvedId and don't collapse into one firing during deduplication.

Tests run locally

• tests/unit/ — 377 passed, 4 skipped (2 browser-smoke failures are chromedriver/env, unrelated)
• tests/compliance/ — 13 passed
• dash-renderer Karma — 25 passed (includes 9 new cases for validation + resolvedId uniquification)
• New Selenium: test_cbwc009_match_input_fixed_output, test_cbwc010_match_input_no_output, test_dvcv017_match_input_permitted_no_output_match — passed
• Regression check: test_cbwc002_fibonacci_app, test_cbwc003_same_keys, test_cbwc008_running_match, test_dvcv006_inconsistent_wildcards — all still pass

Test plan

• CI green
• Spot-check a user app that hit the original error

[T4rk1n]

Looks good 💃