Clean up Java implementation of Thomas Algorithm

[Butt4cak3]

When merged, this PR fixes a few issues with the Java implementation of the Thomas algorithm and also changes it so that it works similarly to the new Julia implementation (#495).

• The class and file name were thomas, but classes in Java should be PascalCase, so I renamed them to Thomas.

• The thomasAlgorithm method had a int size parameter, which is not needed in Java. I assume that the code was directly ported from the C implementation and that was not taken out.

• I had a short talk with @leios on stream today. I mentioned how almost none of the implementations return the result. Instead, they all modify the input vector, which they get passed by reference. The new version of this does not modify any of the input vectors and returns the output vector.

• The output was quite messy, also as a result of the fact that the code was directly ported over from C.

• All the array declarations were formatted as double x[] instead of double[] x. The latter is, as far as I know, the preferred way in Java. This is probably also an artifact of the port from C.

• Lastly, I made the output dynamic. The output is hard-coded in almost all implementations and I think we should change that. It's not that difficult.

Note: I don't know why GitHub reports thomas.py as deleted and Thomas.py as new. I used git mv to rename the file.

[jiegillet]

I support returning a result over modifying the input. I've been requesting that feature on the Thomas algorithm PRs that I reviewed.

[berquist]

Note: I don't know why GitHub reports thomas.py as deleted and Thomas.py as new. I used git mv to rename the file.

If the move also contains enough of a percentage changed over some threshold (can't remember the number), Git views it as a delete and new, rather than a rename.

[PudottaPommin]

Hey, looks good and works like charm.

I have only one question (in code).
I mark this for now as Request changes.