Rework to_batch_interval_start for the new Time world

[jcjones]

This is just not ergonomic. I did it this way because to_batch_interval_start() is just odd to reason through hanging off of DateTimeExt -- for this purpose anyway. It's also a bit weird to hang off Time because then we need basically two time_precision arguments.

I'm going to open a follow-on to tackle this better. For now, this may be dumb but it works.

Originally posted by @jcjones in #4219

[tgeoghegan]

As of right now, we only have a single usage of DateTimeExt::to_batch_interval_start left in the codebase, in an integration test. So instead of reworking that method, I think we should probably remove it. Diving into the diff from #4219 (here's the merge commit; the delta is too big for GitHub's PR view 1) suggests that we should be able to use DateTimeExt::to_time in that one spot.

What remains and is pertinent to this issue is three places where we do something like:

Time::from_time_precision_units(
    (report.client_timestamp().as_time_precision_units() / batch_window_units) 
        * batch_window_units
)

...to align a number of time precision units to a bucket size, itself in time precision units (2). e.g., we might have a report timestamp of 17 time units, but a batch time window size of 4, in which case the timestamp gets "aligned" to 16.

DateTimeExt::to_batch_interval_start(&self, time_precision: &TimePrecision) -> Result<Self, Error> is IMO not the right interface for this because (1) everywhere we do this, we do it on a Time, not a DateTime and (2) we're not aligning to a time precision. We're aligning to something like a second order time precision. So I think we want TimeExt::to_batch_interval_start(&self, batch_interval_duration: Duration) -> Result<Self, Error>.

One other thing falls out of looking at the #4219 diff: usages of DateTimeExt::to_batch_interval_state were replaced with either DateTimeExt::to_time or sometimes Clock::now_aligned_to_precision. The two are very similar, and both wind up using Time::from_seconds_since_epoch and relying on the plain old / operator to round timestamps down. I think we should have just one way to do it, and replace usage of clock.now_aligned_to_precision(&precision) with clock.now().to_time(&precision).

[tgeoghegan]

#4400 deals with to_batch_interval_start. #4401 removes Clock::now_aligned_to_precision.