fix(dogstatsd): reject tag_length.end() <= MIN_TAG_LENGTH upfront

[blt]

Summary

The dogstatsd::common::tags::Generator wrapper subtracts one from tag_length.end() to account for the on-wire : separator, then forwards the adjusted range to common::tags::Generator. The inner generator requires start() >= MIN_TAG_LENGTH = 3. When the caller's tag_length.end() equalled MIN_TAG_LENGTH -- e.g. ConfRange::Constant(3) or ConfRange::Inclusive { min: 3, max: 3 } -- the wrapper produced Inclusive { min: 3, max: 2 }. The inner then rejected the range on min <= max and the failure bubbled up as Error::StringGenerate, which drops the underlying message and points callers at pool generation rather than at the tag-length range that was actually at fault.

This change validates the constraint at the wrapper entrypoint and returns a dedicated Error::TagLengthEndTooSmall variant naming the offending value and the minimum required. The wrapper docstring is updated to record the post-adjust end > MIN_TAG_LENGTH contract.

Discovered while fuzzing DogStatsD::new from an external rig: the symptom was a StringGenerate failure at specific seeds, with no indication that the actual root cause was the configured tag-length range collapsing to its lower bound.

Test plan

• cargo test -p lading-payload -- 211 tests pass including three new wrapper tests (tag_length_constant_at_min_rejected, tag_length_inclusive_min_equals_max_at_min_rejected, tag_length_at_min_plus_one_accepted).
• cargo clippy -p lading-payload --all-targets -- clean.
• CHANGELOG entry added under Unreleased.