Avoid per-argument work when coercing literal arguments

[patrick91]

For context, codspeed flagged some performance regression here, so I asked Claude to investigate and found a quick win 😊

Stacked on #276, which adds the test_execute_field_arguments_sync benchmark used below. Until #276 is merged this PR also shows that commit; only the last commit (Avoid per-argument work when coercing literal arguments) belongs here.

This removes two bits of work that ran for every argument of every executed field:

1. coerce_argument built its default-value error handler on every call. The on_arg_default_value_error closure is only needed when an argument falls back to its default value, but it was created before checking that. It's now built by a small _default_value_error_handler helper, only in the two branches that apply a default value.
2. coerce_input_literal called replace_variables for every leaf literal. A bare variable is already handled at the top of the function, so by the time we reach a leaf type only list and object literals can still contain variables. Other literals (like 2 in value(multiplier: 2)) are already constant, so they're now passed to the scalar's coerce_input_literal directly.

Behavior is unchanged: the full test suite and the doctests of both modules pass, as do ruff check src tests, ruff format --check src tests and mypy src tests.

Locally, test_execute_field_arguments_sync from #276 gets 2–4% faster (median of 200+ rounds, alternating before/after over 4 runs), with about 3 fewer function calls per field. It's a modest win, but it's on the hot path for any query that passes arguments.

AI: I've done this with Claude 😊

[codspeed]

Merging this PR will not alter performance

✅ 19 untouched benchmarks
🆕 1 new benchmark

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
🆕	test_execute_field_arguments_sync	N/A	120.6 ms	N/A

---

_{Comparing patrick91:optimize-argument-coercion (bdab43b) with main (894141f)}

[Cito]

Thank you @patrick91, will have a look as soon as I find some time. Is this ready for review (you marked it as draft)?

[patrick91]

Thank you @patrick91, will have a look as soon as I find some time. Is this ready for review (you marked it as draft)?

I wanted to double check it when it wasn't late as a night :D and also make sure the benchmark was good 😊