refactor(get_lrc, get_node): use numpy methods, add examples

[mwtoews]

This PR replaces internal processing of get_lrc and get_node to use NumPy's unravel_index and ravel_multi_index. There is a minor speed bump, and the methods also check the bounds of the inputs, raising ValueError if they are outside the L/R/C/node range.

Also add docstring examples and minor formatting to adhere to the numpydoc style guide.

[codecov]

Codecov Report

Merging #1591 (3b53949) into develop (2eeefd4) will decrease coverage by 0.0%.
The diff coverage is 100.0%.

@@            Coverage Diff            @@
##           develop   #1591     +/-   ##
=========================================
- Coverage     72.6%   72.6%   -0.1%     
=========================================
  Files          251     251             
  Lines        54419   54413      -6     
=========================================
- Hits         39553   39547      -6     
  Misses       14866   14866             

Impacted Files	Coverage Δ
flopy/discretization/structuredgrid.py	46.8% <100.0%> (-0.4%)	⬇️

[christianlangevin]

This is nice, @mwtoews. Great to use numpy routines directly. Thanks.

I wonder if we should move away from get_lrc() and call it get_cellid() instead (or maybe have both for now), where cellid is (layer, row, column) for a structured grid, (layer, cellnumber) for a vertex grid, and (nodenumber) for an unstructured grid? The reverse can be implemented for get_node().

[wpbonelli]

@langevin-usgs with this plan I think get_cellid would then be equivalent to get_lni for vertex grids — noting here to remind myself to add a test case if this change is made, to make sure calling them on a vertex grid returns the same thing