Trial handling: additional functionality and documentation

[kohlerca]

This PR is a revision of the existing Trials objects. Several improvements are implemented:

Functionality

1. Added additional methods to retrieve spike train or analog signal data across the trial repetitions (get_spiketrains_trial_by_trial and get_analogsignals_trial_by_trial).
2. Moved redundant code to the base class Trials, for improved maintenance.
3. Refactored unit tests to include content checks. The tests are based on a base unit test class that is shared across different test cases and provides custom assertions to compare Neo objects.
4. Changed name convention of arguments to identify trials as indexes (trial_index), instead of the previous trial_id or trial_number.

Documentation

1. Module docstring text improved.
2. General revision of all classes and methods docstrings, to harmonize style and add detailed content of the API.
3. Added usage examples to TrialsFromBlock and TrialsFromList.

TODOs

• Sphinx configuration to handle links to Neo classes in the documentation.

[coveralls]

coverage: 88.323% (+0.06%) from 88.262%
when pulling 6f018b8 on INM-6:enh/trial_by_trial_methods
into 5392583 on NeuralEnsemble:master.

[CozySocksAlways]

While working on instantaneous_rate fix PR #649 , I noticed that sphinx doesn't create a link for :class:Trials . Further check on RTD seems to reveal there is no page for this base class? dedicated pages only exist for the subclasses (e.g. TrailsFromBlocks etc..).
Should we leave this as is, add a page for the base class, or link to the existing Trials documentation wherever a reference is needed(e.g. in statistics.py etc..)?

[CozySocksAlways]

One fix i tried out is using the :mod: cross referencing, so this takes us to the right module page. So for trials one lands on the trials page, from there TrailsFromBlocks or the other one can be selected.