feat(payload): adopt Probability for Config::sampling_probability

[goxberry]

What does this PR do?

Switches Config::sampling_probability (in lading_payload::dogstatsd) from
bare f32 to Probability (BoundedProbability<{ f32::to_bits(0.0) }> from
#1876). The try_from impl enforces finite + [0.0, 1.0] at deserialize
time, so the matching range/is_finite() check in Config::valid() is
removed. The new type threads through MemberGenerator::new and
MetricGenerator::new; at the comparison site in
<MetricGenerator as Generator>::generate, .get() extracts the inner
f32 so RNG draws and bit-exact output are preserved.

Motivation

Phase 1, PR 3 of the stacked rollout planned in #1876 — wire
BoundedProbability aliases into lading_payload configs one field at a
time, smallest blast radius first. After TimestampConfig::probability
(#1877) and ValueConf::float_probability (#1878),
Config::sampling_probability is next: it threads through the
MemberGenerator::new -> MetricGenerator::new constructor chain in
dogstatsd.rs / dogstatsd/metric.rs, with one hot-path comparison site.

Moving the validation into the type means an invalid
sampling_probability is rejected at parse time with a structured error
instead of inside Config::valid(); the in-code field can no longer hold
a value outside [0.0, 1.0].

Related issues

Stacked on #1878 (PR 2: ValueConf::float_probability), which sits atop
#1877 (PR 1) and #1876 (Phase 0). Remaining fields in this rollout:
Config::multivalue_pack_probability (PR 4, Probability) and
Config::unique_tag_ratio (PR 5, AtLeastOneHundredth).

Additional Notes

• Wire format unchanged. Probability serializes as a bare f32 via
  serde(into = "f32", try_from = "f32"), so YAML/JSON configs written
  for the previous field type continue to parse.

• Public API breakage at the Config struct literal. The
  sampling_probability field is pub; external callers that construct
  Config literally (rather than via Default) must now wrap the value
  in Probability::try_new(...). The in-tree default and all test
  fixtures use ..Default::default() for this field, so no in-tree
  fixtures need updating.

• No sampler swap. The hot-path comparison stays
  OpenClosed01.sample(rng) < sampling_probability.get(); .get() is a
  const fn returning a copied scalar and inlines away.

• Default value preserved. Default for Config now constructs
  Probability::try_new(0.5).expect("0.5 is in [0.0, 1.0]") in place of
  the bare 0.5.

• Verified. cargo build --workspace, cargo clippy -p lading-payload --all-targets -- -D warnings, and cargo test -p lading-payload --lib
  (250 pass) all clean.

• Bench delta. cargo bench -p lading-payload --bench dogstatsd --baseline pr2 (baseline saved from feat(payload): adopt Probability for ValueConf::float_probability #1878's tip):

  • dogstatsd_setup: −0.19% (p = 0.59, no change)
  • dogstatsd_throughput/1 MiB: −18% point estimate, but CI is
    [−36%, −3.3%] (p = 0.13, no change — noisy bench at this size)
  • dogstatsd_throughput/10 MiB: −1.7% (p < 0.05, criterion classifies
    as "Change within noise threshold")
  • dogstatsd_throughput/100 MiB: −0.53% (p = 0.39, no change)
  • dogstatsd_throughput/1 GiB: −0.27% (p = 0.66, no change)

  The corresponding --bench block run against pr2 shows scattered
  ±2-3% "regressions" on cache_setup and cache_advance plus a −9.0%
  "improvement" on cache_read_at/1048576, on code that this PR does
  not touch. That sets a wide machine noise floor and confirms the
  dogstatsd_throughput deltas are run-to-run variance, not a real
  effect — consistent with the expectation in the rollout plan (".get()
  inlines away").

[goxberry]

Warning

This pull request is not mergeable via GitHub because a downstack PR is open. Once all requirements are satisfied, merge this PR as a stack on Graphite.
Learn more

• feat(payload): adopt AtLeastOneHundredth for Config::unique_tag_ratio #1881
• feat(payload): adopt Probability for Config::multivalue_pack_probability #1880
• feat(payload): adopt Probability for Config::sampling_probability #1879 👈 (View in Graphite)
• feat(payload): adopt Probability for ValueConf::float_probability #1878
• feat(payload): adopt Probability for TimestampConfig::probability #1877
• feat(payload): improve BoundedProbability ergonomics for adoption #1876
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[datadog-datadog-prod-us1-2]

✨ Fix all issues with BitsAI

⚠️ Warnings

🚦 1 Pipeline job failed

Changelog Check | changelog-check     

🛟 This job is unlikely to succeed on retry. Please review your pipeline configuration.

No changes to CHANGELOG.md detected. Add 'no-changelog' label if this is intentional.

Useful? React with 👍 / 👎

_{This comment will be updated automatically if new data arrives.
🔗 Commit SHA: 7d6eabc | Docs | Datadog PR Page | Give us feedback!}