[Feature]: Add composable SerDes transformation pipelines

[zhongkechen]

What would you like?

Add a composable SerDes processing pipeline while preserving the existing object-to-string SerDes[T] API as the value-codec boundary.

Today, transformations such as filesystem offloading, compression, encryption, signing, or custom envelopes must each wrap and own another complete SerDes. This makes independent transformations difficult to reuse or reorder.

The proposed model is:

value -> root SerDes -> string stage -> string stage -> checkpoint
value <- root SerDes <- string stage <- string stage <- checkpoint

Possible Implementation

• Introduce a synchronous SerDesStage protocol whose serialize/deserialize methods accept and return str plus SerDesContext.
• Add a ComposableSerDes or composition helper that starts with one existing SerDes[T], applies stages in forward order during serialization, and reverses them during deserialization.
• Require stages to emit a self-identifying/versioned format. During deserialization, a stage should:
  • reverse valid input in its own format;
  • raise SerDesError for malformed input that identifies itself as that stage;
  • return unrecognized input unchanged.
• Keep existing SerDes implementations and configuration APIs backward compatible.
• Allow FileSystemSerDes functionality to be exposed as a stage so callers can choose any root value codec and can place compression, encryption, or signing before or after storage as appropriate.
• Add tests for stage order, pass-through of external invocation input, null/empty values, malformed recognized envelopes, and custom value types.

Is this a breaking change?

No.

Does this require an RFC?

Yes.

Additional Context

Python's current FileSystemSerDes already supports a configurable inner value SerDes. A pipeline would separate that value-codec responsibility from reusable string transformations instead of adding another nested-wrapper layer. A comparable Java design is being developed in aws/aws-durable-execution-sdk-java#648.