TimedeltaArray freq validation without _from_sequence

[jorisvandenbossche]

This aligns TimedeltaArray.__init__ with DatetimeArray.__init__, using the same approach how @jbrockmendel now handled the freq validation in DatetimeArray in the merged #24686
First commits removes usage of from_sequence (from #24666), second commit adds the freq validation as in #24686

So related to the discussion at the end #24666 (and a partial revert of that). This makes the TimedeltaArray constructor again more restricted (only accepts correctly typed (int or timedelta64) containers.
Until we decide on the constructors in general (#24684), I would prefer to keep them strict: it is always easier to later expand functionality (and then also test it), than remove functionality.

[codecov]

Codecov Report

Merging #24723 into master will decrease coverage by 49.31%.
The diff coverage is 64.7%.

@@             Coverage Diff             @@
##           master   #24723       +/-   ##
===========================================
- Coverage   92.38%   43.07%   -49.32%     
===========================================
  Files         166      166               
  Lines       52321    52344       +23     
===========================================
- Hits        48337    22545    -25792     
- Misses       3984    29799    +25815

Flag	Coverage Δ
#multiple	?
#single	43.07% <64.7%> (ø)	⬆️

Impacted Files	Coverage Δ
pandas/core/arrays/timedeltas.py	41.25% <64.7%> (-46.61%)	⬇️
pandas/io/formats/latex.py	0% <0%> (-100%)	⬇️
pandas/core/categorical.py	0% <0%> (-100%)	⬇️
pandas/io/sas/sas_constants.py	0% <0%> (-100%)	⬇️
pandas/tseries/plotting.py	0% <0%> (-100%)	⬇️
pandas/tseries/converter.py	0% <0%> (-100%)	⬇️
pandas/io/formats/html.py	0% <0%> (-99.35%)	⬇️
pandas/core/groupby/categorical.py	0% <0%> (-95.46%)	⬇️
pandas/io/sas/sas7bdat.py	0% <0%> (-91.17%)	⬇️
pandas/io/sas/sas_xport.py	0% <0%> (-90.15%)	⬇️
... and 124 more

---

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 d46d2a5...19c5160. Read the comment docs.

[codecov]

Codecov Report

Merging #24723 into master will increase coverage by <.01%.
The diff coverage is 94.11%.

@@            Coverage Diff             @@
##           master   #24723      +/-   ##
==========================================
+ Coverage   92.38%   92.38%   +<.01%     
==========================================
  Files         166      166              
  Lines       52321    52344      +23     
==========================================
+ Hits        48337    48360      +23     
  Misses       3984     3984

Flag	Coverage Δ
#multiple	90.81% <94.11%> (ø)	⬆️
#single	43.07% <64.7%> (ø)	⬆️

Impacted Files	Coverage Δ
pandas/core/arrays/timedeltas.py	88.21% <94.11%> (+0.35%)	⬆️
pandas/util/testing.py	88.09% <0%> (+0.09%)	⬆️

---

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 d46d2a5...19c5160. Read the comment docs.

[TomAugspurger]

Doing this for the RC?

[jorisvandenbossche]

I would personally do that, but it might need some discussion with @jreback / @jbrockmendel ?

[TomAugspurger]

Yeah. Though I'm also comfortable with doing this after the RC is tagged, but before 0.24.0.

[jreback]

looks fine. @jbrockmendel

[TomAugspurger]

I think let's just do this now. Apologies for ramming this through @jbrockmendel.

[jbrockmendel]

No sweat; I'm eager to see the RC too.

This does re-add a lot of redundant code. If the only objection to the previous implementation was that it let too many dtypes through, then just disallow those in the same line that disallowed np.bool_

[jorisvandenbossche]

This does re-add a lot of redundant code. If the only objection to the previous implementation was that it let too many dtypes through, then just disallow those in the same line that disallowed np.bool_

Looking back at the discussion, I think other concerns were performance / the sequence_to_td64 doing too much other work ?