add inplace and batched quadgk

[lxvm]

Fixes #14

This pr also adds inplace and batched QuadGKJL, and it removes all special cases for certain algorithms in tests, since those issues appear to be fixed. I did have to fix one of the test functions that wasn't behaving correctly for 1d batched problems, since it's unclear whether the input points of a 1d batched problem should be a 1xN Matrix or a N-element vector. The docs suggest it is the former, however for most algorithms it is the latter. I took the interpretation that the last axis of the input point array is the "batching" dimension

[lxvm]

It seems like the VEGAS tests are failing for integrand 2, which is a cosine. So the Monte Carlo is probably failing because of a sign problem. I should probably either disable the test or add some constant offset to the integrand to mitigate the sign issue

[ChrisRackauckas]

MonteCarloIntegration.jl is just kind of bad. We should probably update the docs to recommend against using it and mark it as failing. I think there's an issue with its implementation but I haven't found the time to dig in.

[lxvm]

Oddly enough, the tests pass for me locally and using the same versions of everything. I'll revert the VEGAS test I added back to see if that helps

[lxvm]

I just decided to remove the VEGAS test, which should take some of the randomness out of the tests