Mesh partition

[kevindugan]

I'm also interested in the feature outlined in this issue. I've put together a starting point for the associated wrapper.

Some features which may be missing are:

1. The ability to supply a connectivity consistent with the METIS format (ie flat connectivity and offset vector). Initial implementation only supports list of [list of node indices]; the METIS format is deduced.
2. Further options handling in the wrapper. Initial implementation uses mostly default options.

Happy to discuss any suggestions.

[inducer]

Hmm, odd coinidence. Take a look at #44, which seems to pursue goals relatively similar goals to this PR (IIUC that is). Could the two of you come up with a joint proposal so that we can pursue that? Thanks in advance!

[inducer]

Thanks! A few comments below.

[kevindugan]

@inducer, I think this is ready to merge. I incorporated your suggestions, then ran the checks locally and all passed.

[inducer]

Thanks! This is almost good. Just one more thing as far as I can tell.

[inducer]

LGTM, thanks!