Evaluate fixed precision C tests

[drinckes]

This is a test PR only - not for merging. It relates to #669.

This changes the C encoding test to only test the fixed precision (integer) encoding implementation.

It splits the fixed precision part of encoding in the C implementation into a separate function, and modifies the encoding tests to directly test it. It also adds a test runner for MacOS to see if we can still get encoding differences.

[drinckes]

A previous PR that added a macos-latest test runner encountered C encoding failures with (among others):

59.32,-157.21,12,93F48QCR+2222

Instead of the desired code, it generated 93F48QCQ+2X55. This must be down to some very small difference in the conversion of the longitude (since the difference is on even digits "Q" and "X"). macos-latest uses arm64/M1 (ubuntu hardware isn't specified.)

Changing the tests to use fixed precision lat/lngs (computed on ubuntu) gives us:

3733000000,186695680,12,93F48QCR+2222

Running the tests on macos-latest all pass. So this is one way for us to have reliable cross-platform tests.

UPDATE: I mean yes the tests fail but that's because of the formatting check. The encoding unit tests pass.