[MAINT] ASR: defensive hardening and docstring fixes

[sappelhoff]

A group of small defensive and documentation fixes in the ASR module (no behavior change on the common paths):

• asr_process: read state by key (state["M"], state["T"], state["R"]) instead of state.values(), removing the reliance on dict insertion order.
• fit_eeg_distribution: nudge the descending gridsearch arange stop endpoint so min_width is not dropped by floating-point error (mirrors the existing cols grid nudge).
• ASR: expose max_bad_chans as a constructor parameter (default 0.3, preserving current behavior) and document it.
• clean_windows: build the window-start offsets from the non-truncated win_len * sfreq to avoid ~1-sample phase drift for non-integer products (byte-identical when the product is integer, including all existing tests); fix the win_len docstring default (0.5, not 1); document the meegkit-specific flat-window removal criterion.

Testing

Adds tests for the max_bad_chans parameter, state key-order independence, and the offset spacing. Full tests/test_asr.py passes; ruff clean.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 83.10%. Comparing base (1fcfbfd) to head (389a184).

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #117   +/-   ##
=======================================
  Coverage   83.10%   83.10%           
=======================================
  Files          25       25           
  Lines        2817     2818    +1     
=======================================
+ Hits         2341     2342    +1     
  Misses        476      476           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[sappelhoff]

Note: includes the nonlinear_eigenspace reproducibility fix from #118. test_asr_class[*-riemann] compares incremental-vs-bulk output with a loose tolerance and is flaky on CI because that path used an unseeded RNG; this can trip on any PR regardless of what it changes. #118 is the root fix — once it merges first, this commit drops out on rebase.