BUG: Evoked.get_data() should exclude bad channels when explicit picks are given

[evasafi]

Fixes #12577

Epochs.get_data(picks="eeg") excludes channels marked as bad by default,
but Evoked.get_data(picks="eeg") does not. Same method name, same
library, inconsistent behavior.

Root cause: Evoked.get_data() always calls
_picks_to_idx(self.info, picks, "all", exclude=()), unconditionally
passing exclude=(). Epochs._get_data() only does this when picks
is None; when explicit picks are given, it lets _picks_to_idx fall
back to its default of exclude="bads".

This PR mirrors that branching logic in Evoked.get_data():

• picks=None : unchanged, bads still included (preserves existing behavior)
• explicit picks (e.g. picks="eeg"): bads now excluded by default,
  matching Epochs.get_data()

Added a regression test in test_evoked.py::test_get_data that fails
on current main and passes with this fix. Also added a changelog
entry and added myself to the contributor list.

[CarinaFo]

Hi Eva, thank you for your contribution.
Fully agree with you that the inconsistent behaviour between epochs and evokeds should be fixed.
I am not sure if changing the behaviour of evokeds is the way to go, it might break people's code if we don't properly deprecate that.
Additionally, it seems that this is a broader and more complex issue as mentioned here.

Let me discuss that with the other maintainers and get back to you.

[evasafi]

Thanks for the context, Carina. That forum thread and the broader consistency question make sense. Happy to keep this PR scoped to the Evoked.get_data() fix + docstring for now, and adjust the approach (deprecation path, etc.) once you've discussed with the other maintainers. If it'd be useful, I'd also be glad to help think through the broader .pick() standardization once there's a direction. That's a discussion I'd been meaning to follow more closely anyway.