Exodus: I/O of time series data, field data, cell data

[Intergalactyc]

Overview

Fairly significant overhaul to _exodus.py to provide the following features:

• Reading of time series data: optional timestep kwarg can either be an integer (default 0) to select a given timestep by index, or "all" to instead obtain a list with a Mesh for each timestep. Improved read performance by pre-fetching variable handles and disabling NetCDF auto-masking where not yet done.
• Writing of time series data: new optional kwargs for the write function to manually specify point_data and cell_data (list of dicts ordered by time) with multiple timesteps, as well as time array parameters (and time inference when these are not completely specified)
• Reading and writing of field data (global variables)
• Splitting of multidimensional arrays (vectors) into scalar components with uppercase suffixes (X, Y, Z) during write, and symmetric re-categorization via the (fixed to better accomodate 2d) categorize helper during read to reconstruct original meshio vector shapes; update of existing vector handling during read to work on cell data just as will point data
• Writing of cell data (reading was already supported, but writing was not) with multi-block consistency.
• Fix to element block indexing, as SEACAS based tools often require 1-indexed element blocks
• Writing of side sets (for now only when manually specified)
  Minor changes were also made to _helpers.py to allow the additional timestep kwarg for read to be passed through (non-breaking).

Backwards compatibility

Fully backwards compatible. All changes made are nonbreaking: call signature changes are only the addition of optional kwargs, and return values should be the same unless these kwargs are used.

Tests

Added four new tests to test_exodus.py to comprehensively test new features. These and all existing exodus tests pass.

Relevant issues

• exodus time-steps #381 (and linked [REQUEST] Support exodus time series #1328): this provides the general time series support desired
• Node sets and side sets input/output #249: not a complete solution for this (I didn't want to extend the already wide scope of this too much more), but the manual side-set specification added to write can easily be adapted to full support for side sets as described. Node sets are already supported.

Anyways, thanks @nschloe and everyone for this great tool; hopefully these changes can help someone.