Remove per-row index level across components

[philsmt]

This is an API breaking change to correct one of my really stupid ideas.

The first components using PulsePattern components to label data added another index level for every individual row, e.g. hitIndex for hits, signalIndex for signals etc. This was meant to have an easy way to look into the first hit of every pulse, e.g. At the same time, this vastly complicates indexing pulse-resolved data with each other as pandas trips over the additional index level. Without this level, it's straightforward to index pulse-resolved data both with a train-resolved index as well as pulse-resolved index. The original functionality of getting the first hit for each pulse can easily be achieved using .groupby() if needed.

I also stopped using this level in Timepix3, so this change now fixes this for DelayLineDetector and AdqRawChannel.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 58.20%. Comparing base (95c59de) to head (45c2c21).
⚠️ Report is 182 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #362      +/-   ##
==========================================
+ Coverage   57.87%   58.20%   +0.32%     
==========================================
  Files          31       31              
  Lines        4905     4962      +57     
==========================================
+ Hits         2839     2888      +49     
- Misses       2066     2074       +8     

☔ View full report in Codecov by Sentry.
📢 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.

[takluyver]

I guess we can assume that not too many people have started using this code and relying on the extra index so far, so if there's any breakage beyond our own code, it should be minimal. 🤞

[philsmt]

I want to be honest here, this code has quite definitely been used over the past run in most applicable experiments. It is being advertised by example notebooks. I have not seen any use of this index level however, which coupled with the fact it makes indexing to trains/pulses harder is the reason I want to get rid of it. I waited for the end of the cycle/LIMP, so the 2025/1 environment remains compatible in case anyone relies on it.

[takluyver]

OK, fair enough. I think we need to be careful with backwards compatibility, but it doesn't need to be 100% absolute, and it sounds like the benefits here outweigh the drawbacks. 👍

[philsmt]

Breaking compatibility is the intent here.

Maybe I should add a direct note to the method's documentation?

[takluyver]

Sorry, I wasn't writing very clearly. I understand that this breaks compatibility, and I think it's an acceptable trade-off. 👍

[philsmt]

@takluyver Was that an LGTM? 🙃

[takluyver]

LGTM

[philsmt]

I decided to defer merging this MR towards the end of the cycle, to prevent the small off chance to break someone's analysis in the middle of a cycle.