Mean 1d

[aradi]

• Substantially reduced code & number of fypp-constructs in mean().
• Extended mean() to handle reduction from 1D-array to scalar (similar to sum())
• Introduced tolerances when testing floating point results (note checking with == 0.0 is highly unreliable and may give false positives)
• Reformatted test file according to style guide.

Commits in order of relevance, feel free to cherry pick if you don't like all of them.

[jvdp1]

These improvements look really good to me (I tried to achieve a similar thing but without success; the reduce_shape is really useful). Thanks @aradi .

[certik]

Can you please submit the formatting changes in a separate PR? Let's not pollute our PRs with formatting.

Btw, the style guide says to use 4 spaces:

https://github.com/fortran-lang/stdlib/blob/11002bd4797b0fe015bf831a53f8794e32ea32b4/STYLE_GUIDE.md

Either way, that is irrelevant to the actual changes in this PR, which all look good to me. So if you could please split the formatting into a new PR, then let's get it PR merged.

[aradi]

Well, the styleguide says the body of every Fortran construct should be indented by two (4) spaces, so I have chosen the one I prefer more 😆 .

Anyway, I have taken out the reformatting commit, it really does not belong to here, I agree. I'll wait with reformatting changes until the style guide becomes non-ambigous. 😉

[certik]

@aradi ha, that was a mistake, thanks for spotting it. I fixed it in #131.

[certik]

@aradi the PR looks great, thank you!