Gracefully handle missing availability service

[lkeyson-earthscope]

The FDSN fdsnws-availability service and MUSTANG's ts_* time-series metrics (e.g. ts_percent_availability_total, ts_num_gaps) are currently down for an undetermined period. Though their return will likely occur at the same time, these two services could go down and come back independently of each other. This PR lets PIQQA reports keep generating while either or both are unavailable, and switch back onto them automatically (no manual flag/config change) once they recover.

What changed
Independent health checks, one per service, run once per report:

• doAvailability() probes fdsnws-availability directly (PIQQA.py) and only treats it as down on an explicit HTTP 410 (or another HTTP error / connection failure) - no longer relies on a getcode() == 410 check that could never actually be reached (urlopen raises HTTPError for 4xx/5xx rather than returning a response object).
• reportUtils.checkTsMetricsExist() (new) independently probes MUSTANG for ts_percent_availability_total against a stable reference station (IU.ANMO), so we don't conflate "availability service is down" with "ts_ metrics are down" - they're detected and handled separately.

Fallback paths when a service is down:

• reportUtils.get_extents_from_mustang() (new) computes data start/end extents from MUSTANG's percent_availability measurements when fdsnws-availability is unreachable, replacing the availability-service lookup used to determine data start/end for PDF and spectrogram plots. If it cannot retrieve percent_availability metrics, it will use the report dates.
• reportUtils.getDataStartEnd() (new) wraps the "get start/end for a target" logic (previously duplicated 4x across doPDFs/doSpectrograms) and picks between getAvailability() and get_extents_from_mustang() based on the availability-service flag.
• doAvailability() skips straight to the percent_availability fallback (instead of attempting ts_percent_availability_total and only falling back after a failed retry-with-sleep) when checkTsMetricsExist() has already determined ts_ metrics are down — avoids ~5-10s of wasted retries per channel group.
• doBoxPlots() drops ts_num_gaps_total and substitutes num_gaps for ts_num_gaps in the requested metric list when ts_ metrics are down.

Updated urls:

• service.iris.edu --> service.earthscope.org

This MR also bumps python to 3.12, off of end-of-life 3.9.

Example report that was run on the latest code:
UW_2026.zip

[gillian-earthscope]

Mostly just comments about updating IRIS references to EarthScope