Running multi-threaded and handling warnings

When using the CLI to process an orthomosaic the program will as default run as multi-threaded. Because of a open bug in numpy (https://github.com/numpy/numpy/issues/27989) calculating the color distance which is using numpy matmul the program will crash on windows.

To avoid this problem we use a lock when doing the color distance calculation. This solve the problem, but at the cost of a small performance penalty. Even with the lock we still see a large improvement in performance compared to running as a single thread.

Another problem solved by introducing the lock on color distance calculation was numpy giving RuntimeWarnings on invalid inputs in np.sqrt leading to invalid data in the generated color distance orthomosaic. The cause of this is unknown, but is assumed to originate from simultaneously reading from the same array when calling calculate_distance from the ColorModel class.

Since RuntimeWarnings can lead to invalid output, and users are not expected to react to warnings, all warnings will be handled as errors. If a specific warning needs to be handled differently this can be done by setting filterwarnings specifically for that warning in __init__.py. For example:

warnings.filterwarnings("ignore", category=NotGeoreferencedWarning)