Refactor hitlets

[WenzDaniel]

What is the problem / what does the code in this PR do
With this PR we refactor the hitlet data_type a bit.

1. get_hitlet_data also supports now empty inputs. A corresponding test was added.
2. The the hitlet dtype with data field was extended by goodness of split and max gap. This allows to share the _split_peaks function of our PeakSplitter class. Hence we can remove some redundant code.
3. I refactored straxen's hitlet plugin by adding a new function called create_hitlets_from_hits. This gives the plugin a bit more structure and generalizes hitlets such that it can also be used more easily for other analyses. Further, this allowed to remove the refresh_hit_to_hitlets.

This PR goes together with #463.

[JoranAngevaare]

In principle this looks very good, I think all was already mentioned in #435 and #430.

Given our ordeal in #309, would you mind double checking that no annoying edge-cases are occurring when processing hitlets after peaks and vise versa since the dtypes might be slightly different?

[WenzDaniel]

Given our ordeal in #309, would you mind double checking that no annoying edge-cases are occurring when processing hitlets after peaks and vise versa since the dtypes might be slightly different?

I checked only for hitlets. But good point I should check for both.

[WenzDaniel]

Okay processing looks good.