fix: timestamp parsing tests think they're changing the timezone but they aren't

Noticed here: https://github.com/posit-dev/connectapi/pull/405/files#diff-0d56414e1bc981e4745c47aaf583ffd17cc170be5b80ee3ff31bf056e543903aR134

According to ?Sys.timezone, the timezone is detected the first time it is needed and then cached for the rest of the R session. So setting TZ within the R process isn't actually effective.

% R
> Sys.timezone()
[1] "America/New_York"
> options(TZ="UTC")
> Sys.timezone()
[1] "America/New_York"

We missed that in the tests because we were asserting that the strings were being parsed according to Sys.timezone(), and we thought that was changing by setting TZ, but it wasn't.

It's probably not worth doing anything fancier to prove that we set different timezones, so we should probably just delete the duplicate tests that suggest that we can change the timezone and keep the ones that assert that we read in Sys.timezone().