Enable the use of Caffeine as an alternative to OpenCoarrays

[rouson]

• Add Caffeine as a dependency in the fpm manifest (fpm.toml)
• Change the extension in the name of files that will need preprocessing: .f90 -> .F90
• Add use caffeine_m statements using the C preprocessor
• Condition the compilation of the new use statements in Matcha on a C preprocessor macro such as USE_CAFFEINE. For example, the matcha_m module will need

#ifdef USE_CAFFEINE
    use caffeine_m, only : this_image => caf_this_image
#endif

• Repeat the 2nd through 4th steps above for the Caffeine dependency Sourcery, focusing on the data_partition class. If many additional parts of Sourcery need similar revision, we might consider splitting data_partition_t off from Sourcery to become its own library.
• Also repeat the 2nd through 4th steps above for the Sourcery dependency Assert.
• Add instructions to the README.md for parallel execution with Caffeine: probably something like fpm test --flag "-DUSE_CAFFEINE"

[everythingfunctional]

@Dominick99 , @rouson and I had a brief discussion about how to handle the circular dependency between assert and caffeine. caffeine will "vendor" (keep its own copy of) the assert module. This code will then conditionally use the caffeine assert. I.e.

#ifdef USE_CAFFEINE
    use caffeine_assert_m, only: assert
#else
    use assert_m, only: assert
#endif

I'll throw together a PR for caffeine this afternoon, and then you can take care of using it here. Let me know if you have any questions.