Update storage router to write blobs when reading

[epociask]

Problem
Currently secondary storage locations are only utilized after dispersing a blob through proxy. This doesn't compensate for reading or verifier/syncing nodes who operate irrespective of the batcher. Furthermore the operation of writing to secondary storage targets increases response latency.

Proposed Solution
Add a secondary storage routine which transits blobs seen either via dispersal or retrieval from EigenDA to cache and fallback targets if and only if they haven't been persisted to the secondary storage target before.

[samlaf]

Thinking through this: is this feature actually useful? Derivation only needs to read this once right? Unless there’s a reorg or a challenge. But is it worth paying the extra write cost in the happy path for these rare cases?