Qslib

- `Protocol` now has `Protocol.stage`, and Stage now has `Stage.step`, to provide convenient, 1-indexed access,
such that `protocol.stage[5]` of is stage 5 of `protocol`, not stage 6.
- `Protocol` now supports setting PRERUN and POSTRUN stages, as a series of SCPI commands. This allows
the setting of things like idling temperatures and exposure times. It is not easily usable yet, however.
- `Experiment.change_protocol_from_now` allows convenient changes to a currently-running experiment.
- `Normalization` has been renamed to `Processor`. Plotting functions can take sequences of processors to
process data. These now include:
- Normalization as before: `NormByMeanPerWell`, `NormByMaxPerWell`
- `SubtractMeanPerWell`: subtracts the mean of a particular region, applied to each well.
- `SmoothEMWMean`: smooths data using Pandas' ExponentialMovingWindow.mean.
- `SmoothWindowMean`: smooths data using Pandas' Rolling.mean or Window.mean.
- Some initial implementation changes to allow repeated steps (not stages).
- Fixes bug that prevented loading of some aborted runs.
- Fixes monitor's recording of temperatures.
- Experiment.all_filters uses data if it exists; Experiment.filter_strings as a convenience function alternative.
- SCPICommand parsing improvements.
- Protocol printing improvements.

- Fixes drawer check bug.

- Add checks for cover and drawer position after changing positions.

- Fix pyparsing 3.0.7 whitespace parsing problem (pyparsing/pyparsing359).
- Use Hypothesis for some tests.

- Improved plots, including new `Experiment.plot_temperatures`.
- Comments in SCPI commands, allowing save/load of default filters in protocols
- Licensing switched to AGPL 3.0, CLA in docs.
- Command comment parsing, also used to store whether steps are using default filters.
- Example notebook
- Working parsing of Exposure commands
- Dependency version fixes

- `CustomStep.collect` (and subclasses, including `Step`) is now `CustomStep.collects`. The `collect` parameter for
`Step` can now be `None`, which will collect data if the step has a filter setting (ie, if you want to collect the
default `Protocol` filters, you need to use `collect=True`).
- `Stage.stepped_ramp` and `Stage.hold_for` are now safer in requiring some keyword arguments, are documented, and
are more flexible.