WiP: ReproIn heuristic - support usecase from @mih et al.

[yarikoptic]

• allow/strip any leading CAPITALS followed by a column with the series/protocol name (see xnat support ReproNim/reproin#14)
• consider series_description after protocol_name if it was not only empty, but didn't parse according to our spec (thus should "allow" for protocol_name to be a fancy "xnatDownload" as in xnat support ReproNim/reproin#14)

[codecov-io]

Codecov Report

Merging #187 into master will decrease coverage by 0.07%.
The diff coverage is 78.33%.

@@            Coverage Diff             @@
##           master     #187      +/-   ##
==========================================
- Coverage   68.33%   68.26%   -0.08%     
==========================================
  Files          30       30              
  Lines        2334     2357      +23     
==========================================
+ Hits         1595     1609      +14     
- Misses        739      748       +9

Impacted Files	Coverage Δ
heudiconv/bids.py	81.52% <60%> (-1.92%)	⬇️
heudiconv/convert.py	77.13% <80%> (-0.15%)	⬇️
heudiconv/heuristics/reproin.py	83.7% <82.22%> (-0.51%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 4eff1a6...41ec485. Read the comment docs.

[mih]

If it wouldn't be 'WiP' I would have merged it in order to get a little further with datalad/datalad-neuroimaging#17

[yarikoptic]

bloody python3 zealots ...

SecretStorage requires Python '>=3.5' but the running Python is 2.7.14

so we could expect lots of fun with deployments of datalad if pip would not become smart to not even consider "new" versions for incompatible python. According to PLOS/allofplos#20 it should be the case for pip >= 9. I think we get 10, and I guess that doesn't feature doesn't work since secret storage does have
python_requires='>=3.5', , so either virtualenv causes older pip to be used somehow (yet to check) or that feature in pip is buggy

[mih]

we could expect lots of fun with deployments of datalad

Mission accomplished: datalad/datalad#2442

[yarikoptic]

ok, tests passed and I hope that this doesn't ruin anything, merging

[mih]

FTR: With this PR, but without ContentDate/Time:

WARNING: Failed to get date/time for the content: Dataset does not have attribute 'ContentDate'.
Traceback (most recent call last):
  File "/home/mih/env/datalad3-dev/bin/heudiconv", line 11, in <module>
    load_entry_point('heudiconv', 'console_scripts', 'heudiconv')()
  File "/home/mih/hacking/heudiconv/heudiconv/cli/run.py", line 120, in main
    process_args(args)
  File "/home/mih/hacking/heudiconv/heudiconv/cli/run.py", line 330, in process_args
    overwrite=args.overwrite,)
  File "/home/mih/hacking/heudiconv/heudiconv/convert.py", line 198, in prep_conversion
    overwrite=overwrite,)
  File "/home/mih/hacking/heudiconv/heudiconv/convert.py", line 292, in convert
    save_scans_key(item, bids_outfiles)
  File "/home/mih/hacking/heudiconv/heudiconv/bids.py", line 237, in save_scans_key
    op.join(output_dir, 'sub-{0}{1}_scans.tsv'.format(subj, ses)), rows)
  File "/home/mih/hacking/heudiconv/heudiconv/bids.py", line 270, in add_rows_to_scans_keys_file
    data_rows_sorted = sorted(data_rows, key=lambda x: (x[1], x[0]))
TypeError: '<' not supported between instances of 'NoneType' and 'str'