Cached heuristic name discrepancy

[mgxd]

Summary

When reconverting and using a different heuristic filename from the cache, the conversion table (filegroup.json) is still reused. This results in heudiconv ignoring the new heuristic and defaulting to the grouping from the previous heuristic. Relevant code excerpt:

heudiconv/heudiconv/convert.py

Lines 125 to 133 in 312cd2e

	target_heuristic_filename = op.join(idir, op.basename(heuristic.filename))
	# TODO:
	# 1. add a test
	# 2. possibly extract into a dedicated function for easier logic flow here
	# and a dedicated unittest
	if (op.exists(target_heuristic_filename) and
	file_md5sum(target_heuristic_filename) != file_md5sum(heuristic.filename)):
	# remake conversion table
	reuse_conversion_table = False

...and saving...

heudiconv/heudiconv/convert.py

Line 154 in 312cd2e

safe_copyfile(heuristic.filename, idir)

Proposed solution - save used heuristic in cache simply as heuristic.py to avoid any naming discrepancies.