With just density, the l=2,4 multipoles are extremely small, and the GSL
integrator would abort the computation, so now I just turn it off for both
`compute_corrfunc` and `compute_multipole`.

For now, it only has the `covariance_matrix` function, but I'm sure more will
be added.

It's useless on its own anyway, and may confuse users if they accidentally
instantiate it.

It requires the entirety of scipy as a dependency, so it's best to put it in a
separate module.

According to PEP 518, one needs to declare Cython in this file, instead of in
setup.py.

Hopefully everything is being changed consistently now.

Mostly modified some explicit casting, and it all works.

Now the signature of the functions is available when using Coffe interactively,
or when generating the docs.

This is to be able to add the `covariance_matrix` method, which returns the
block-diagonal matrix that can be used for, say, Fisher forecasts.

Now instead of taking a function, they take two lists, one z, the other
bias(z). This is to prevent double interpolation (one by Python sampling a
non-analytic function, one by GSL), and it's probably more familiar to the
user.

I should make it possible to set either of those

I don't want to figure out how distutils packaging works just yet, for now it's
easiest to put everything in one enormous file and package that.

Forgot to set the power spectrum each time the contributions changed

This one was quite challenging to find; apparently, by calling
`set_power_spectrum_linear`, I modified k_min and k_max, but whenever I
re-computed it using CLASS, it was not being adjusted accordingly, so the
integrals would all give `nan`s, which of course propagated further and
caused the tests to fail. Hopefully, that's all I missed, seems to work now.