[feat] Add configurable jitter strategy (`full`, `equal`, `decorrelated`, `none`) to backoff calculations

[AryanSharma48]

Description:

Problem

Currently:

• smooth-api-ts hardcodes Equal Jitter (half + Math.random() * half).
• smooth-api-py hardcodes Full Jitter (random.uniform(0, capped)).
  Users cannot configure deterministic backoff for testing or choose their preferred jitter distribution algorithm.

Proposed Solution

1. Add jitter?: 'full' | 'equal' | 'decorrelated' | 'none' to BackoffConfig in both TypeScript and Python.
2. Implement standard AWS-style jitter algorithms:
   • none: Deterministic exponential ($base \times 2^{attempt}$) capped at $maxDelay$.
   • full: Random value between $0$ and $capped$.
   • equal: $half + \text{rand}(0, half)$.
   • decorrelated: $min(maxDelay, \text{rand}(baseDelay, prevDelay \times 3))$.
3. Default to equal across both packages for parity.

Acceptance Criteria

• BackoffConfig updated in packages/smooth-api-ts/src/types.ts and packages/smooth-api-py/smooth_api/config.py.
• calculateBackoff / calculate_backoff updated in both packages.
• Unit tests added verifying that calculated delays stay strictly within [0, maxDelay] bounds for all 4 strategies.

References

• TypeScript Implementation
  Add configurable jitter strategies to backoff calculation in TS #53

• Python Implementation
  N\A

[shreyansh0207]

Hi @AryanSharma48, apologies — I forked the repo and implemented this directly instead of commenting first to ask. I've opened a PR with the TypeScript implementation: #53

Summary of what I did:

• Added jitter field to BackoffConfig in types.ts
• Implemented all 4 strategies (none, full, equal, decorrelated) in calculateBackoff
• Added prevDelay tracking in the retry loop so decorrelated jitter works correctly
• Added unit tests covering all 4 strategies, verifying delays stay within [0, maxDelay]

I focused on TypeScript only for now — happy to implement the Python side (smooth-api-py) as well if you'd like, just let me know! And let me know if you'd prefer I follow a different process next time.

[AryanSharma48]

@shreyansh0207 No worries! Just a good thing to get assigned first as it removes possibility of 2 people working on the same issue!

The implementation was good, ive left some comments regarding some nitpicks. Have a look at them and you can proceed with the python implementation as well!