Source parsing tests

[LKedward]

As discussed in #155, this uses the error_t type to return and propagate errors from the new source parsing routines.
Adds a test suite for the parsing routines - any difficult parsing edge cases can be added in there as they are found.

There are other regions from #155 that still require test coverage.

[awvwgk]

Looks good. I have one minor suggestion for the file_parse_error. Since we have the line number already, it might be more verbose to print the line of the source file as well, maybe, in case we know, even highlight the offending part in some way.

I did something similar for the TOML parsing with a context from the tokenizer to provide some visual aid together with the error message:
https://github.com/toml-f/toml-f/blob/d7371fc8/src/tomlf/error.f90#L215-L243

[LKedward]

Thanks @awvwgk 👍

.. it might be more verbose to print the line of the source file as well, maybe, in case we know, even highlight the offending part in some way.

Great idea! That's a neat implementation too!

I've now added more context output, though in most cases indicating the offending column may have to wait until a proper parsing library is used.

[certik]

We should move these into an fpm_string.f90 module, and eventually get these into stdlib.

But let's first merge this PR, then we can refactor later.

[everythingfunctional]

This looks great. I know writing tests is often a thankless chore, but it looks like you've really shored this part up with it, so thanks! :)