reserveCells crashes mint mesh inside Umpire

[gunney1]

Crash reserving space for UnstructuredMesh

• This PR is a bug fix
• It does the following:
  • Adds tests that verify the crash (which happens on our docker platforms)
  • Fix the bug (if it is indeed a bug) or make the code die gracefully

This crash has been plaguing PR #1263. The MarchingCubes test crashes while asking mint::UnstructuredMesh to reserve space. This PR aims to provide a reproducer and correct the crash.

Two identical tests were added:

• In src/axom/quest/tests/quest_initialize.cpp
• In src/axom/mint/tests/mint_mesh.cpp

Interesting behaviors:

• The test added to quest reproduces the crash, but the one added to mint does not.
• The crashes only occur in our docker builds.
• Bypassing the memory reserve calls doesn't work for the originating problem. The crash occurs later, when memory allocation takes place.

[gunney1]

Some new observations:

1. Based on @white238's guess, I disabled MPI (and MFEM, which was configured with MPI) from the failing config. This fixed the failure. Not exactly sure what the mechanism of failure was.
2. I tried to construct the UnstructuredMesh without using Sidre. This also gets past the crash. This provides a work-around for this bug, but it effectively disables some Sidre-dependent features when using the UnstructuredMesh.

[gunney1]

After merging in the latest develop branch, the failed test (quest_initialize) continues to fail. But if I remove the -DBUILD_SHARED_LIBS=ON from the configuration, the tests pass. @white238 What do you think of this clue?