Fix 219

[tomaskontrimas]

Apply mask to events_meta as we initially draw events from all datasets and signal_event_post_sampling_processing expects the same shape of events_meta and events arrays.

Fixes #219 issue. It does not change the signal injection behaviour when events fall within valid event field ranges (e.g. declination range after events rotation). Trials would fail when generating signal for analyses with multi datasets and/or multi source hypothesis groups.

It follows the same logic as in the initial signal events drawing in L886-897:

skyllh/skyllh/core/signal_generator.py

Lines 886 to 897 in f8e4ea6

	ds_shg_mask = ds_mask & shg_mask
	shg_sig_events_meta = sig_events_meta[ds_shg_mask]
	n_shg_sig_events = len(shg_sig_events_meta)
	ev_idx = shg_sig_events_meta['ev_idx']
	# Get the signal MC events of the current dataset and source
	# hypothesis group.
	shg_sig_events = mc[ev_idx]
	# Do the signal event post sampling processing.
	shg_sig_events = shg.sig_gen_method.\
	signal_event_post_sampling_processing(
	shg, shg_sig_events_meta, shg_sig_events)

[tomaskontrimas]

Oops, I guess analysis unit tests https://github.com/icecube/i3skyllh/actions/runs/8241075808 fail due to the livetime calculation change before the v24.1.0 release (see the test on master/v24.1.0 https://github.com/icecube/i3skyllh/actions/runs/8241246057/job/22538113710).

Let's fix them first, before merging this fix.

[martwo]

Good catch! Thanks for fixing this.