Re-implement .trials

[tensionhead]

Current Situation

The backing Indexer class is a sophisitcated, yet quite inefficient iterable, which gets re-computed on every .trials property access. Meaning that if someone just accesses repeatedly my_data.trials[1000], all preceding 1000 trials get consumed via _get_trial, exhausting the 'computed' generator, every single time. This is sub-optimal for continuous data, but apparantly really abysmal for discrete data, see also #403

Idea

A simple fixed/'pre-computed' .trials property subclassed from list which maps item access to _get_trial

def __getitem__(self, idx):
    return self._get_trial(idx)

should do the trick. Support for slices, arrays etc. can then be added as we have it right now..

[tensionhead]

Done with #422