Merge TCN, noise resiliency option, addition of more DIII-D shots, etc. from forked FRNN

[felker]

Mostly located on @ge-dong's forked version of the code.

• [x] Should the additional D3D shots be directly added to the existing list, or should it be kept separate for posterity with the Nature paper results? Kept separate, in both shot_lists/.txt files (d3d_clear_data_avail.txt vs.d3d_clear_since_2016.txt, e.g.) and internal variables in conf_parser.py (d3d_full vs. d3d_full_2019)
  • [ ] shot_lists/d3d_clear.txt and shot_lists/d3d_disrupt.txt are currently unused. Names of all files in shot_lists/ are confusing. See below for details.
• [ ] Re-tune the hyperparameters for d3d_data and d3d_data_0D datasets on the expanded shot list.

(base) ➜  shot_lists wc -l d3d_clear*
  10000 d3d_clear_10000.txt
   1000 d3d_clear_1000.txt
    100 d3d_clear_100.txt
  18122 d3d_clear_data_avail.txt
  27311 d3d_clear_since_2016.txt
  74248 d3d_clear.txt
 130781 total
(base) ➜  shot_lists head -n1 d3d_clear*
==> d3d_clear_10000.txt <==
143800   -1.000000

==> d3d_clear_1000.txt <==
165500   -1.000000

==> d3d_clear_100.txt <==
167475   -1.000000

==> d3d_clear_data_avail.txt <==
125502   -1.000000

==> d3d_clear_since_2016.txt <==
125502   -1.000000

==> d3d_clear.txt <==
40560   -1.000000
(base) ➜  shot_lists tail -n1 d3d_clear*
==> d3d_clear_10000.txt <==
168554   -1.000000

==> d3d_clear_1000.txt <==
168554   -1.000000

==> d3d_clear_100.txt <==
168554   -1.000000

==> d3d_clear_data_avail.txt <==
168554   -1.000000

==> d3d_clear_since_2016.txt <==
180844   -1.000000

==> d3d_clear.txt <==
168554   -1.000000

[felker]

Closed via #50 and #49.