pd.Timestamp.utcfromtimestamp should return tzaware

[jbrockmendel]

ATM pd.Timestamp.utcnow() returns a UTC-localized Timestamp object, but pd.Timestamp.utcfromtimestamp() does not. As a result, the following round-trip fails:

unow = pd.Timestamp.utcnow()
result = pd.Timestamp.utcfromtimestamp(unow.timestamp())
assert unow == result  # <-- TypeError because unow is tzaware and result is tznaive

Changing this would mean breaking with the stdlib, but our utcnow already breaks with the stdlib (see also #22450)

[mroeschke]

+1 I think breaking from the stdlib is fine since we offset rich support for timezones and we do not advertise Timestamp as a direct replacement for datetime.datetime

[jbrockmendel]

BTW, there is now a thread on the datetime stdlib SIG mailing list (https://mail.python.org/archives/list/[email protected]/thread/PT4JWJLYBE5R2QASVBPZLHH37ULJQR43/) about changing datetime.utcnow and datetime.utcfromtimestamp behavior