change import name to mumps and project to python-mumps

Closes #8

See merge request !10

Closes #8

switch to meson build

Closes #11 and #19

See merge request kwant/mumpy!9

Add a wrapper for the MUMPS null space algorithm

Closes #7

See merge request kwant/mumpy!5

Closes #4

See merge request kwant/mumpy!7

Add tests for all datatypes.

Closes #2

See merge request kwant/mumpy!4

Using a large, dense and completely random matrix with a sparse
solver takes a long time, but this test is only testing whether
MUMPSContext correctly increases the memory limit when it detects
that MUMPS has run out. This means we can choose a matrix that
is sparse and easy to factorize, as long as the memory requirements
are large.