WIP: modify PyDarshan plotting routines to accept records as input

[Yanlilyu]

• This branch modifies the existing plotting routines to take records as input rather than report objects. Three plotting files are changed, which are plot_opcounts.py, plot_access_histogram.py, and plot_common_access_table.py.

• plot_opcounts() supports records input of POSIX, MPIIO, and STDIO modules, but doesn’t provide supports for those of H5F, H5D, PNETCDF_FILE, and PNETCDF_VAR modules. In order to sustain these modules, codes related to c library need to be revised. For now, if users try to generate plots for H5F, H5D, PNETCDF_FILE, PNETCDF_VAR modules, an error will be raised. Codes related to “agg_ioops” are deleted.

• plot_access_histogram() supports records input of POSIX and MPIIO. It is not supporting HDF5 and PNETCDF modules. We will add the functions later. Codes related to “mod_agg_iohist” are deleted.

• plot_common_access_table() accepts records input of POSIX, MPIIO modules.

• We added xfails for some tests using HDF5/PNETCDF and updated the test xticks and labels to test_plot_exp_common.py.

[tylerjereddy]

I suppose one other suggestion I might have on the testing side is to compose expected "accumulated" results from the individual expected values in i.e., darshan/tests/test_plot_exp_common.py. For example, read in STDIO from ior_hdf5_example.darshan and dxt.darshan, and require plot_opcounts to produce bar heights that are the sum of the two current expected values in test_bar_heights.