Improve read_raw_eyelink

[scott-huberty]

Describe the new feature or enhancement

Per discussion with @larsoner (see this comment), while I am focused on the eye-tracking project this summer, it may be a good idea to take another pass at mne.io.read_raw_eyelink.

Describe your proposed implementation

a review of read_raw_eyelink can be focused on:

1. increasing code coverage
2. identifying code that can be refactored to be more idiomatic to mne-python
3. identifying bottlenecks where performance can be improved

Describe possible alternatives

Open to suggestions.

Additional context

I will try to pick away at this during the down time between other tasks for my GSoC project.