Assorted set of fixes related to DateTime Z formatter

[rouault]

I'm perhaps missing something but it seems Qt cannot handle the Z formatter and doesn't automatically perform conversion to UTC when the input datetime is in another time spec.

[github-actions]

🪟 Windows builds ready!

Windows builds of this PR are available for testing here. Debug symbols for this build are available here.

(Built from commit 7a2c34c)

[nyalldawson]

These incremental changes look good to me.

(I do think we need a bigger rework of time zones at some stage, eg having some method to allow the utc datetime values in a GPKG to always be shown to the user in local time. )

[nyalldawson]

@rouault should we backport this?

[rouault]

@rouault should we backport this?

I'm kind of nervous, as I'm not totally sure this will not cause backward compatibility problems.
For ex in #57262, the user specified a default value as format_date(datetime_from_epoch(epoch(now())+3600000*4),'yyyy-MM-ddTHH:mm:ssZ'), with probably the intent to shift its local time by 4 hours and have that computed value directly considered as UTC. Now this will convert that computed value (its local time + 4 hours) to UTC.

The commit in the OGR provider should be OK to backport.