Store calibration and other metadata in well-documented form

[saeub]

Description of the problem

After merging #884, some metadata are still left in the metadata dictionary (stored as a private attribute GazeDataFrame._metadata). In particular:

• Calibrations
• Validations
• Blinks
• Changes in tracked eye or sampling rate (see feat: add support for multiple recording specs per file to gaze.from_asc() #887)

There should be a well-documented public interface to access these metadata.

Description of a solution

I see several options to approach a solution:

• Storing those metadata in custom data structures as attributes of GazeDataFrame (similar to the metadata stored in Experiment, EyeTracker etc.)
• Storing time-dependent meta-events (like calibrations, validations, blinks) in a DataFrame similar to fixations and saccades (but it's probably difficult to find a useful set of columns for all the different types of events)

We might need to think about what metadata is available for non-EyeLink devices too, and how they could all fit into the same data structure.

Minimum acceptance criteria

• Public interface to access (ideally) all metadata returned from parse_eyelink() (and possibly other file formats)
• Documentation for every type of metadata
• Remove GazeDataFrame._metadata attribute (see feat: Store metadata from ASC in experiment metadata #884 (comment))

[dkrako]

Probably the easiest next step will be to pass the parsed blinks as an EventDataFrame to the GazeDataFrame.__init__().
The dataframe will need the three columns: name, onset, offset. The name should be blink.