Remove a function composition from totientSum by using a better choice for triangle

[incertia]

This also better matches the formula described in mathworld.
On the other hand, totientSum = (`div` 2) . (+1) . generalInversion ((^2) . fromIntegral) is another candidate by following the formula from wikipedia, which kills a where clause as well.

The latter might offer very minor performance improvements due to shuffling around of arithmetic operations.

I chose to keep using the triangle inversion formula because that was the previously used formula.

[cartazio]

pardon my possible confusion, but what does this commit achieve? could you walk me through your reasoning a bit more?

[Bodigrim]

Well, it is a small, but still valuable improvement IMO. I am for it. To make it even more valuable, @incertia, could you please also add case totientSum n = 0 for non-positive n (instead of internal error, which is current behaviour!) and reflect this change in comment and in tests?

I'd rather keep version with triangle, and add a link to the source of formula in comment.

[incertia]

all it really does is avoid a composition with (+1) in the definition of totientSum and match the formula in the first search result for "euler totient sum" to avoid any possible confusion if someone examines the source and doesn't recognize the formula used

apart from that, not much.

[incertia]

I made the changes @Bodigrim mentioned as well as added two special test cases (totientSum 0 should be 0, and totientSum (negative number) should be 0). If you guys are ready to merge the changes I can squash the commits.

[cartazio]

oh ok, i read the diff wrong, i thought you had added the (+1) . :)))))

my bad!

i leave @Bodigrim to make the call as he sees fit :)

[Bodigrim]

@incertia please use haddock markup for URLs.

[incertia]

I added angle brackets around the URLs and force pushed a squash.

[Bodigrim]

Great! Thanks for your contribution.