Document snapshot copies, snapshot storage, and origin-first HTTP caching (#2286) * docs: explain snapshot copies, snapshot storage, and origin-first caching Acceleration snapshots are full copies on standard object storage, separate from Cayenne S3 Express One Zone. HTTP caching with a zero TTL serves the accelerator only after the origin fails. Co-authored-by: Luke Kim <lukekim@users.noreply.github.com> * docs: fix the snapshots link on the Spicepod reference and quote zero cache durations as 0s The Spicepod reference linked ../../features/data-acceleration/snapshots from reference/spicepod/index.md, one ../ too many for an index page. The 2.3.x build failed on it, and trunk resolved to the 2.3.x page. A bare caching_ttl: 0 is a YAML integer, and the runtime rejects it: Invalid 'caching_ttl' param value: Int(0). Expected a duration string. Write the zero durations in the origin-first examples as 0s. * docs: scope origin-first caching to the accelerator and keep the cold tier on S3 The SQL results cache answers an identical query within its 1s item_ttl, including a fallback served during an outage, before the request reaches the accelerator or the origin. Say so on the caching-mode and read-through pages, with Cache-Control: no-cache and sql_results.enabled: false as the ways to send every query to the origin. An accelerated HTTP dataset skips the connector's response cache, so origin max-age plays no part. The Cayenne cold tier takes only a general-purpose S3 or S3-compatible bucket. The performance-tuning summary row and three nearby sentences grouped it with the GCS and ADLS snapshot backends. --------- Co-authored-by: Cursor Agent <cursoragent@cursor.com> Co-authored-by: Luke Kim <lukekim@users.noreply.github.com>