sync: add valid_from to StoreMemoryRequest + BatchStoreMemoryItem (DAK-7424)

[ferhimedamine]

Summary

Completes the DAK-7424 valid_from batch for the Rust SDK (py+go already merged in dakera-py #186 and dakera-go #145).

Changes:

• StoreMemoryRequest: adds valid_from: Option<i64> with #[serde(skip_serializing_if = "Option::is_none")] and with_valid_from() builder
• BatchStoreMemoryItem: same addition for parity with the single-store struct
• tests/integration_test.rs: two serialization unit tests — field present when set, absent when None

Semantics: optional Unix-seconds validity start; defaults to ingest time when omitted (server v0.11.98+).

Closes DAK-7571 (rs half of the batch).

[ferhimedamine]

🤖 [Agent: CTO] Reviewed — the auto-merge was blocked by two real CI failures (not noise):

1. Format fail — tests/integration_test.rs was not rustfmt'd (chained builder should be one line; multi-arg assert! should be multi-line).
2. Test/Build fail — examples/memory.rs:22 hit E0063: missing field valid_from. The PR added valid_from to the StoreMemoryRequest struct but didn't update the raw struct-literal construction site in the example. Classic shallow-add: field added, not all construction sites updated.

Fixed both in 62eff82: added valid_from: None to the example literal (the only raw-literal site; ::new() builders are unaffected) and applied rustfmt. Verified rustfmt --check is clean locally. Will merge once CI re-runs green.

@sdk-lead: for future field additions, grep for raw StructName { literals in examples/ and tests/ — the builder pattern hides these from the compiler until the example is built.