feat(payload): adopt Probability for ValueConf::float_probability

[goxberry]

What does this PR do?

Switches ValueConf::float_probability 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, and the public
constructor ValueConf::new now takes the validated type directly.
(ValueConf::valid() already had no finite/range check on this field, so
nothing is removed there.) The new type threads through
NumValueGenerator's Constant and Uniform variants; at the two
comparison sites in <NumValueGenerator as Generator>::generate, .get()
extracts the inner f32 so RNG draws and bit-exact output are preserved.

Motivation

Phase 1, PR 2 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), ValueConf::float_probability is the next smallest: it touches
only dogstatsd.rs and dogstatsd/common.rs, and ValueConf is
constructed via Default or ValueConf::new (no test fixtures build it
with a struct literal), so the change stays localized.

Moving the validation into the type means an invalid float_probability
is rejected at parse time with a structured error instead of silently
accepted (the existing valid() never checked it); the in-code field can
no longer hold a value outside [0.0, 1.0].

Related issues

Stacked on #1877 (PR 1: TimestampConfig::probability), which itself sits
atop #1876 (Phase 0). Remaining fields in this rollout: Config::sampling_probability,
Config::multivalue_pack_probability, Config::unique_tag_ratio (via
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 ValueConf::new. The constructor signature
  changes from new(float_probability: f32, ...) to
  new(float_probability: Probability, ...). The field is private, so
  external callers that build ValueConf go through new, Default, or
  deserialization. There are no in-tree callers of ValueConf::new.

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

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

• Bench delta. cargo bench -p lading-payload --bench dogstatsd --baseline pr1 (baseline saved from feat(payload): adopt Probability for TimestampConfig::probability #1877's tip):

  • dogstatsd_setup: −0.29% (within 95% CI of zero)
  • dogstatsd_throughput/1 MiB: −5.6%
  • dogstatsd_throughput/10 MiB: −2.5%
  • dogstatsd_throughput/100 MiB: −4.8%
  • dogstatsd_throughput/1 GiB: −5.0%

  All deltas are favorable or neutral. The --bench block run against the
  same pr1 baseline shows ±1-5% swings on cache_* benches that don't
  touch any code in this PR, which sets the noise floor — the
  dogstatsd_throughput "improvements" fall in the same band and are
  consistent with run-to-run variance rather than a real speedup.

[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
• feat(payload): adopt Probability for ValueConf::float_probability #1878 👈 (View in Graphite)
• 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]

✨ 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: 8f98d9a | Docs | Datadog PR Page | Give us feedback!}

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 0c1423ac1f

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Remove the panic path from the default config

The repository instructions for /workspace/lading explicitly say production code must not use .unwrap() or .expect(), but this new default value adds an .expect() in ValueConf::default. Even though the current literal is valid, this leaves a panic path in production initialization and violates the no-panic invariant for config construction; prefer a non-panicking constant/helper for the known-valid default probability.

Useful? React with 👍 / 👎.

[blt]

Hmm hmm hmm same comment as in #1877 (comment) but here it's 50%. This may just be a limit of Rust's type system we have to accept.

[blt]

+1 here with one comment that is shared in the parent. Choice in parent will resolve the nit here.