Fix the bug when passing pixel-registered matrix as grids

[seisman]

Description of proposed changes

Fixes #3502

Reminders

• Correct base branch selected? master for new features, 6.1 for bug fixes
• Make sure that your code follows our style. Use the other functions/files as a basis.
• Add tests for new features or tests that would have caught the bug that you're fixing.
• Describe changes to function behavior and arguments in a comment below the function declaration.
• If adding new functionality, add a detailed description to the documentation and/or an example.

[seisman]

@PaulWessel You can work on this branch directly. No need to merge it into master.

[PaulWessel]

OK, need to wrap another debug first...

[PaulWessel]

The gmtapi_init_matrix confused what the offset argument to gmt_M_get_n should be (it should be the registration, but we passed 0.5 * registration). To add insult to injury, we did not set the matrix registration using the input registration, hence the matrix was always gridline registered.
Good test!. The plots now look the same, @seisman.

[PaulWessel]

I checked that gmt_M_get_n is used correctly everywhere else.

[seisman]

Great! Please add some comments to src/testapi_matrix_as_grid.c, and add a test in the test/api directory.

[PaulWessel]

I've added the test and comments.