Optionally return volume-tag-to-element map from read_gmsh

[majosm]

Possibly supersedes #292.

[majosm]

@inducer, @alexfikl Any ideas how I could appease pylint here?

************* Module meshmode.mesh.io
[926](https://github.com/inducer/meshmode/runs/5873887120?check_suite_focus=true#step:3:926)
meshmode/mesh/io.py:354:14: E1101: Instance of 'tuple' has no 'vertices' member (no-member)
[927](https://github.com/inducer/meshmode/runs/5873887120?check_suite_focus=true#step:3:927)
meshmode/mesh/io.py:356:16: E1101: Instance of 'tuple' has no 'vertices' member (no-member)
[928](https://github.com/inducer/meshmode/runs/5873887120?check_suite_focus=true#step:3:928)
************* Module test_mesh
[929](https://github.com/inducer/meshmode/runs/5873887120?check_suite_focus=true#step:3:929)
test/test_mesh.py:548:12: E1101: Instance of 'tuple' has no 'groups' member (no-member)
[930](https://github.com/inducer/meshmode/runs/5873887120?check_suite_focus=true#step:3:930)
test/test_mesh.py:549:48: E1101: Instance of 'tuple' has no 'vertices' member (no-member)
[931](https://github.com/inducer/meshmode/runs/5873887120?check_suite_focus=true#step:3:931)
test/test_mesh.py:712:26: E1101: Instance of 'tuple' has no 'groups' member (no-member)
[932](https://github.com/inducer/meshmode/runs/5873887120?check_suite_focus=true#step:3:932)
test/test_mesh.py:879:15: E1101: Instance of 'tuple' has no 'groups' member (no-member)
[933](https://github.com/inducer/meshmode/runs/5873887120?check_suite_focus=true#step:3:933)
test/test_mesh.py:886:42: E1101: Instance of 'tuple' has no 'vertices' member (no-member)
[934](https://github.com/inducer/meshmode/runs/5873887120?check_suite_focus=true#step:3:934)
test/test_mesh.py:893:30: E1101: Instance of 'tuple' has no 'vertices' member (no-member)

It doesn't seem to like the fact that the result type of read_gmsh depends on its arguments (it thinks it's always a tuple, when AFAICT it's not.)

[alexfikl]

Any ideas how I could appease pylint here?

As far as I know, pylint isn't smart enough to figure this out, so probably best to just add some ignores :\

[majosm]

Looks like 5d9dcb9 fixed it, somehow...

[majosm]

Seems like this is the way to go; it's cleaner than anything I've been able to come up with otherwise. In some ways it's unsatisfying, because it sidesteps my sense that if we're tagging surfaces inside the mesh, we ought to be tagging volumes as well; but I haven't been able to find a nice way to integrate that into the current code structure. I guess this approach could still be consistent with that, as long as we think about a given Mesh as only ever representing a single volume.

[inducer]

LGTM, thanks!