BUG: Cleanup timedelta offset

[sinhrks]

• closes pd.Timedelta() cannot process offset alias T while pd.infer_freq() return this format. #21762
• tests added / passed
• passes git diff upstream/master -u -- "*.py" | flake8 --diff
• whatsnew entry

Fixed to use same unit parsing logic in Timedelta and to_timedelta. Needs to update after #23264 and #23259.

[pep8speaks]

Hello @sinhrks! Thanks for submitting the PR.

• There are no PEP8 issues in the file pandas/core/tools/timedeltas.py !

• There are no PEP8 issues in the file pandas/tests/scalar/timedelta/test_timedelta.py !

[codecov]

Codecov Report

Merging #23439 into master will decrease coverage by <.01%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master   #23439      +/-   ##
==========================================
- Coverage   92.23%   92.23%   -0.01%     
==========================================
  Files         161      161              
  Lines       51197    51189       -8     
==========================================
- Hits        47220    47212       -8     
  Misses       3977     3977

Flag	Coverage Δ
#multiple	90.61% <100%> (-0.01%)	⬇️
#single	42.26% <100%> (-0.01%)	⬇️

Impacted Files	Coverage Δ
pandas/core/tools/timedeltas.py	98.18% <100%> (-0.24%)	⬇️

---

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 24ab22f...04f3ceb. Read the comment docs.

[jbrockmendel]

Is this cimported by another cython module? If not, doesn’t need to be in the pxd file

[jreback]

test comment, otherwise lgtm.

[jreback]

can you make this a parameter (call it box)

[sinhrks]

Thx, fixed.

[jreback]

thanks @sinhrks

[jorisvandenbossche]

Personally, I think we should make a distinction between which units are accepted for the unit parameter, and which are parsed inside a string.

For parsing a string, it is fine to accept both '15m', '15min', '15minute', '15minutes and '15T'. But for the unit keyword, is it needed to allow this flexibility?
I would say that just the standard units (the numpy ones, as we had now) + maybe the abbreviations as used in our freq strings should be enough no? (so 'm' and 'T')

(and to be clear, the rest of this PR, fixing the inconsistency of what strings are parsed is of course totally fine!)

[sinhrks]

@jorisvandenbossche I kept existing behaviour, but it is better to clarify the policy. It should be consistent with period also.

pd.Period('2011-01', freq='min')
# Period('2011-01-01 00:00', 'T')

pd.Period('2011-01', freq='hour')
# ValueError: Invalid frequency: hour