feat(serdes): Add filesystem serdes for large payload storage

[ayushiahjolia]

Issue #, if available: #479

Description of changes:

• Adds create_filesystem_serdes() - a SerDes implementation that stores serialized values on a durable filesystem (Amazon S3 Files or Amazon EFS mounted to Lambda). This keeps checkpoint sizes small regardless of payload
  size, with optional inline preview for observability.

Features -

• ALWAYS mode: Every value written to file; checkpoint stores file pointer
• OVERFLOW mode: Inline if < 256KB, overflow to file if larger
• URI path encoding: Human-readable paths derived from ARN (<functionName>/<executionName>/<invocationId>/<operationId>.json)
• HASH path encoding: Fixed-length SHA-256 paths for safety
• Preview: Optional inline preview in checkpoint envelope for console/API visibility without reading the full file
• ARN parsing: Compact directory structure from durable execution ARNs

Tests -

• Added unit tests, e2e tests with replay and integration tests.

Next PRs -

• Cloud integration tests (requires S3 Files or EFS mount infrastructure)
• Documentation update in aws-durable-execution-docs repo

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[yaythomas]

this used?

and if not, how come ruff didn't notice, I wonder?

the package's pyproject.toml has [tool.ruff.lint] select = ["TID252"], does this means ruff is configured to enable only TID252 (the absolute-imports rule) and all default checks, including F401, are turned off?

[ayushiahjolia]

Not used and yes, select = ["TID252"] means ruff only enables that single rule. F401 (unused imports) and all other default checks are off. The CI's hatch fmt --check only enforces formatting (via ruff format) and that one lint rule.

I can followup with enabling F401 or all pyflakes checks if we want.