feat: nightly hardening - jittered retry backoff

[mouse-value-add]

Problem

When many requests fail at once, all retries currently occur on identical exponential schedules. This can create synchronized retry spikes against providers and increase cascading failures.

Approach

• Added optional to (default )
• Applied bounded random jitter to each retry sleep ()
• Clamped negative jitter input to for safety
• Added focused async tests for jitter behavior and negative-input hardening

Verification

• Ran: ============================= test session starts ==============================
  platform darwin -- Python 3.9.6, pytest-8.4.2, pluggy-1.6.0
  rootdir: /Users/mouse/.openclaw/workspace/nightly/PromptLens
  configfile: pyproject.toml
  plugins: anyio-4.12.1, asyncio-1.2.0, cov-7.1.0
  asyncio: mode=strict, debug=False, asyncio_default_fixture_loop_scope=None, asyncio_default_test_loop_scope=function
  collected 2 items

tests/test_retry_backoff_jitter.py .. [100%]

================================ tests coverage ================================
_______________ coverage: platform darwin, python 3.9.6-final-0 ________________

Name Stmts Miss Cover

promptlens/init.py 3 0 100%
promptlens/main.py 3 3 0%
promptlens/cli.py 169 169 0%
promptlens/exporters/init.py 6 6 0%
promptlens/exporters/base.py 15 15 0%
promptlens/exporters/csv_exporter.py 26 26 0%
promptlens/exporters/html_exporter.py 38 38 0%
promptlens/exporters/json_exporter.py 16 16 0%
promptlens/exporters/markdown_exporter.py 77 77 0%
promptlens/judges/init.py 3 3 0%
promptlens/judges/base.py 15 15 0%
promptlens/judges/llm_judge.py 71 71 0%
promptlens/judges/parser.py 107 107 0%
promptlens/judges/prompts.py 30 30 0%
promptlens/loaders/init.py 4 4 0%
promptlens/loaders/base.py 14 14 0%
promptlens/loaders/json_loader.py 23 23 0%
promptlens/loaders/yaml_loader.py 34 34 0%
promptlens/models/init.py 4 4 0%
promptlens/models/config.py 41 41 0%
promptlens/models/result.py 60 60 0%
promptlens/models/test_case.py 25 25 0%
promptlens/models/tools.py 74 74 0%
promptlens/providers/init.py 3 3 0%
promptlens/providers/anthropic.py 52 52 0%
promptlens/providers/base.py 23 23 0%
promptlens/providers/factory.py 21 21 0%
promptlens/providers/google.py 47 47 0%
promptlens/providers/http.py 67 67 0%
promptlens/providers/openai.py 57 57 0%
promptlens/providers/you.py 57 57 0%
promptlens/runners/init.py 2 2 0%
promptlens/runners/runner.py 96 96 0%
promptlens/utils/init.py 1 0 100%
promptlens/utils/cost.py 15 15 0%
promptlens/utils/diff.py 25 25 0%
promptlens/utils/retry.py 25 4 84%
promptlens/utils/timing.py 24 24 0%

TOTAL 1373 1348 2%
Coverage HTML written to dir htmlcov
============================== 2 passed in 0.24s ===============================

• Result:

Risks

• Retry timing is now non-deterministic by default, which can affect tests that assume exact sleeps.
• Mitigation: jitter remains small (10% by default), and tests can set deterministic random behavior via monkeypatch.

Rollback Plan

• Revert this PR to restore fixed-delay retry behavior.
• Alternatively, set at call sites to disable jitter without reverting code.