Server-side caching for authoritative timeseries

Is your feature request related to a problem? Please describe.
Pulling period of record data via CDA is very slow any may not even be possible for the public to do without multiple requests. (e.g. this request will only work intermittently for me:

curl -X 'GET' \
  'TEST/cwms-data/timeseries?name=Baldhill_Dam.Flow-Out.Inst.~15Minutes.0.best&office=MVP&begin=1948-08-29T15%3A06%3A08.543Z&end=2025-08-29T15%3A06%3A08.543Z' \
  -H 'accept: application/json;version=2

see https://github.com/USACE/cwms-data-api/issues/1229 for possible reasons why it is intermittent)

Describe the solution you'd like
However, it appears A2W is using some sort of caching to return a similar data call very quickly (about 8 seconds when I tried it).
https://water.usace.army.mil/cda/reporting/providers/mvp/timeseries?name=Baldhill_Dam.Flow-Out.Inst.~15Minutes.0.best&begin=1945-08-22T15:20:10.708Z&end=2025-08-29T15:20:10.708Z

Additional context
I suspect most of the CDA calls will be for the elevation pool, tail, inflow, outflow for projects. If they can be cached somehow it might really increase performance for a lot of uses and it could piggy back on whatever is decided for https://github.com/USACE/cwms-data-api/discussions/956.

Start by reviewing issue #1229 and discussion #956, then compare the CDA and A2W timeseries requests described here. A complete proposal would need to define which authoritative timeseries are cached, how caching interacts with the existing API, and how the performance improvement would be measured.