Event builder should include pulses partially inside the event

[JelleAalbers]

When run on events with very large pulses (e.g. the 800us pulses from run 25 the event builder will create event ranges much shorter than the pulses (unless the left and right extension options are set very high). When you process these datasets with pax, the pulses will be truncated considerably.

We can solve this in several ways:

• The event builder automatically extends the event range to accommodate pulses that are included in it. We don't change the criteria for inclusion (start time must be in the range). This can be done either in the input plugin or the filler-reader.
• The processor extends the events instead. The event ranges stored in the raw data will be meaningless in this case.
• We don't solve it at all, you just have to set left and right extension and cross your fingers.

I think the first option is by far the cleanest. We can think about doing this by default or only in a special mode.

See also #43, #315 and 8d1c693a6dd84542836a1ee8c1e81db2ee1de609.

[JelleAalbers]

I've generalized the issue, since even with short pulses it can happen that a pulse starts just outside the event boundary, but ends within it. These pulses should be included along with the event, but currently aren't. We want the events to be complete snapshots of what the channels do. Even for the first few us this matters (e.g. they are routinely used for accidental coincidence studies in XENON100).