type-fixes + enforcements + contracts for analytical solutions

[mathematicalmichael]

return predicted covariance / posterior covariance always.

[mathematicalmichael]

when formulating the problem as a rank-k update, we have an updated covariance matrix that we can return.

may be good to rename mud_sol_alt to reflect the difference. need to update mud-examples accordingly.

tests failing because you need to search for mdf.mud_sol to find the instances where the contract change breaks things.

[mathematicalmichael]

the approach I ended up taking was defining mud_sol_with_cov (and one for map). The "efficient" mud solution avoids direct computation of the updated covariance.

[mathematicalmichael]

closes #43