Normalize docker_service max_failure_ratio diffs to prevent 0 vs 0.0 drift

[Copilot]

docker_service was planning perpetual in-place updates when rollback_config.max_failure_ratio (and similarly update_config.max_failure_ratio) differed only by string formatting ("0" vs "0.0"). The provider state/read path formats this value as a decimal string, which caused Terraform to detect non-semantic drift.

• Schema diff normalization for ratio fields

  • Added a shared diff suppressor (suppressEquivalentFloatRatio) that parses old/new values as floats and suppresses diffs when numerically equal.
  • Applied to both:
    • update_config.max_failure_ratio
    • rollback_config.max_failure_ratio

• Targeted unit coverage for drift semantics

  • Added service schema tests to verify suppression behavior for equivalent float representations and to ensure real numeric changes still plan correctly.

"max_failure_ratio": {
  Type:             schema.TypeString,
  Default:          "0.0",
  Optional:         true,
  ValidateDiagFunc: validateStringIsFloatRatio(),
  DiffSuppressFunc: suppressEquivalentFloatRatio(),
}