Rename C_D to covariance in ESMDA API

[tommyod]

• Aligning the signatures of ESMDA and SIES. Both now take covariance as an argument.

This PR is just a search and replace.

[Blunde1]

My default was to be against this because there are multiple covariances going around in Kalman-type updates. I therefore wanted to build association into the naming. However, the covariance of the measurement error is the only one that is passed as input, as pointed out, so it is perhaps okay. It makes perhaps also sense to step away from the notation in the 2019 paper of SIES, if other algorithms, like ES-MDA (that have different notation in their respective papers), are implemented. A descriptive naming like covariance over C_D then makes sense. My preference is perhaps still cov_d, but given a good docstring I will be okay.