Georinex

Significant overhauling, deduplicating, robustness improvement for RINEX and CRINEX reading.
Particular improvements for:

* CRINEX 1.x files
* uncompressed and compressed OBS 2.x files

In particular, the speed difference between "slow" and "fast" reading of OBS2 files is now less than 10% and memory usage is reduced proportional to `tlim` if specified.

Please let us know of any problem RINEX files.

For OBS2 and OBS3 files, added ability to read by time interval. That is, ready only every say 5 or 30 or whatever seconds. This can greatly speed up parsing of large files when the full time resolution is not needed.
The option is `interval=` and any of float,int (seconds) or datetime.timedelta can be specified

A convenience function georinex.to_datetime() was added. This converts an xarray.DataArray of times to datetime.datetime ndarray--or scalar if it's only one time.

Continuing the work from v1.6.9, we make RINEX *and* Hatanaka CRINEX reading file-extension agnostic. Some organizations like to name their RINEX/CRINEX in non-standard ways.

This actually made the code simpler (!) and more functionalized.

However, compressed files of any type (.gz, .Z, .zip, etc.) are still decompressed by file extension. If this is an issue, we could consider looking at the file magic number, but I have only in the rarest situations seen where people are not using the standard general compressed file extensions.

Enhance test coverage ~ 94% by parametrizing unit tests

BUGFIX: match RINEX 3 spec for Galileo, only 3 ionospheric correction coefficients

Some users may wish to input `io.StringIO` instead of discrete files, which is now possible for all relevant functions.

* detect RINEX type (obs, nav) based on header first line, not file extension. Despite the file naming convention being standard, it seems inevitable a mistake is made that is trivially detected from the first line of a RINEX file.
* add IRNSS NAV3
* better NAV3 spare field handling, not all files have all spare fields.