Allow GMT_Put_Vector to convert other text items than datetime

[PaulWessel]

See this post for background. While we added support for GMT_DATETIME last year, this PR generalizes the conversion from text to numbers via the type GMT_TEXT. The strings may be either Cartesian numbers, datetime, longitude, or latitude, and we also allow these to be converted to other internal representations than GMT_DOUBLE (e.g., GMT_FLOAT, GMT_LONG, etc.) via logical addition (e.g., pass type as GMT_TEXT|GMT_FLOAT).

I have left the prior scheme of GMT_DATETIME for datetime so we are backwards compatible, but one now only needs to select GMT_TEXT and GMT figures out what you passed, hopefully. We assume each column has the same type and there is no error checking if you pass both time and longitude as part of a single column. I added testapi_putvector.c and calling script apiputvector.sh which demonstrates that it works.

I think a quick test and this can be approved and merged.

Just built this branch and ran the full PyGMT tests, the GMT_DATETIME implementation works like before.

Great, maybe @weiji14 could test Lon lat inputs

As for the new data types, GMT_TEXTLON, GMT_TEXTLAT, and GMT_TEXT, does it mean that PyGMT has to know the data types before passing to GMT?

For example, if the string is 12.34E, PyGMT must check if it's text, longitude, latitude or datetime, and then choose the correct type when calling GMT_Put_Vector. It means more work for PyGMT (and other external wrappers) to detect the data types.

I'm wondering if we can simply pass any strings (e.g., 12.34, 12.34E, 12.34N, 12:30:30, 2020-01-01T01:00:00 to GMT giving a general data type (type=GMT_TEXT) and GMT is very clever to parse them and automatically determine the data type.

We could automatically determine type and give an error if more than one, e.g., both time and latitude since a column must only have one type

@seisman, this PR is now more align with your suggestion.

Pinging @weiji14 and @seisman to consider this PR given it has a test that passes.

It works well with a simple test (see GenericMappingTools/pygmt#975 for the example script and output). Need to do more testing later.

So please keep the PR open for a few more days.

Looks good to me.

@weiji14 Could you please bump the GMT dev version again in the conda-forge channel, so that we can better test the new GMT_TEXT data type.

Ok, I'll do that in a bit