Keep the result of stripping the SE prefix from a decision code

[arpitjain099]

get_description() in openfda/device_clearance/transform.py drops the return value of str.replace:

def get_description(data):
  data = data.replace('SESE', 'SE')
  if len(data) > 2:
    data.replace('SE', '')

The line above it assigns, this one does not, so the SE prefix is never actually removed and the lookup runs against the original string.

decision_codes.csv only holds two-character codes, so a value longer than two characters is the SE prefix plus a real code, which is exactly what the length check is testing for. The effect is that every composite code other than SESE resolves to Unknown:

SESE   -> Substantially Equivalent
SESK   -> Unknown
SESD   -> Unknown
SK     -> Substantially Equivalent - Kit
SD     -> Substantially Equivalent with Drug

SESK and SK are the same clearance decision, so the description is being lost for the prefixed form. That value lands in decision_description on the device clearance records.

The change assigns the result. SESE still collapses on the line above and is unaffected, and a two-character code never enters the branch.

Added openfda/device_clearance/tests/transform_test.py covering the plain, doubled, prefixed and unknown cases. The prefixed case fails on master:

openfda/device_clearance/tests/transform_test.py:20: AssertionError
1 failed, 3 passed

With the fix, python3 -m pytest openfda/device_clearance/tests/transform_test.py is 4 passed, and openfda/tests/common_test.py still passes.