Ensure geographic datasets passed by memory files are seen as geographic

[PaulWessel]

Externals (in this case PyGMT) may pass data into GMT modules. These may be created and data added, and the creator may know the data are geographic and may flag that via the bit-flag GMT_GRID_IS_GEO in GMT_Create_Data. Unfortunately, this was not honored. Furthermore, this concept should not be restricted to grids: The user may know that the created dataset, grid, image, matrix, or vector is geographic and would like to pass that information into the data structure. This PR allows for that while fixing the grid-specific bug. Now,

1. GMT_GRID_IS_GEO is deprecated and we use GMT_DATA_IS_GEO instead for all the above data types (the old enum is kept for backwards compatibility)
2. The bit-flag, if set, calls gmt_set_geographic for the chosen direction (input or output source)
3. When GMT_Create_Data builds the various structures we check on that and add a grdtype flag for grids, images, matrices (same value and meaning as in hidden grid header) and a geographic = 0|1 flag for datasets and vectors
4. When gmt_init_module checks if an input file (if found) is a remote data set (and hence sets geographic) we also check if given a memory file to see if it is known to be geographic.
5. When we call GMT_Read_Data and the sources is a memory file we are now able to process the hidden geographic information.

With these changes, a memory file passed that is known to be geographic will trigger an automatic geographic projection if none had been selected explicitly with -J.

Note: Tested with the original PyGMT problem. I assume it will work similarly for images and grids, as well as data sets and vectors, but there are no such external tests (yet). I also updated the API doc to mention the purpose of the GMT_DATA_IS_DATA flag under GMT_Create_Data.

[seisman]

I can confirm that it solves the initial PyGMT problem.

GMT_GRID_IS_GEO is deprecated and we use GMT_DATA_IS_GEO instead for all the above data types (the old enum is kept for backwards compatibility)

Do you mean that PyGMT should pass GMT_GRID_IS_CARTESIAN for a Cartesian grid, and GMT_DATA_IS_GEO for a geographic grid?

[PaulWessel]

No need to pass GMT_GRID_IS_CARTESIAN which is 0 so I dont think that would work - it is sort of the default. So either GMT_DATA_IS_GEO or nothing. It probably should not have been added to gmt_resources.h but...