[FIX] Erased timezone in $toMillis implementation

[LouisAuneau]

Issue

I noticed when converting an ISO 8601 date with a valid timezone, the timezone component is replaced by UTC, which corrupts the actual timestamp.

Reproduction:

from jsonata import Jsonata

expr = Jsonata('$fromMillis($toMillis("2025-01-01T01:00:00+01:00"))')
print(expr.evaluate({}))

You get: 2025-01-01T01:00:00.000Z
While the expected value should be: 2025-01-01T00:00:00.000Z

Proposed solution

Just adding a check to see if the extracted date is timezone aware before using dt.replace(tzinfo=datetime.timezone.utc).

Do not hesitate if you have question or suggestions for some changes. Thank you in advance!

---

PS: Note none of the tests in jsonata base repository use another timezone than UTC, hence tests couldn't catch this case. I think it would be worth contributing to those tests as well. However the JS implementation does handle timezone changes correctly.

[rayokota]

Thanks @LouisAuneau , I've fixed the issue in an alternate PR #27

[LouisAuneau]

Thanks for you quick response @rayokota. However:

1. Converting the timezone aware datetimes to UTC is inefficient as .timestamp() handles all timezones correctly. I personally use jsonata-python to process thousands of entries, so any extra CPU-loop can have an impact.
2. Copying contributed code into your own PR is not really in the spirit of the open-source community...

[rayokota]

Thanks for you quick response @rayokota. However:

1. Converting the timezone aware datetimes to UTC is inefficient as .timestamp() handles all timezones correctly. I personally use jsonata-python to process thousands of entries, so any extra CPU-loop can have an impact.
2. Copying contributed code into your own PR is not really in the spirit of the open-source community...

Apologies, I made the changes before you fixed your PR. In any case, you're right about the unnecessary code, so I've merged your PR.

[LouisAuneau]

Thanks a lot, much appreciated, and thanks for this library, it helps a lot!