Problem

Since TRANSLATE-5226, the number of parallel batch requests for AI pre-translation is configurable per translate5AI language resource in the UI (model property batchParallelRequests, falling back to the global runtimeOptions.LanguageResources.Pretranslation.batchParallelRequests config when unset).

For TQE (AI quality estimation) no such setting exists: the number of parallel requests is hardcoded to 10 in the TQE worker and cannot be changed at all, neither globally nor per resource. Different models and API accounts have very different rate limits - a resource behind a low-tier API key can run into rate limiting with 10 concurrent TQE requests, while a high-tier account could safely run more.

Solution

• Add a separate parallel requests for TQE setting to the translate5AI model properties form in the UI, analog to the existing Parallel batch requests when pretranslating field (batchParallelRequests) - separate from the pre-translation value, because rate-limit pressure differs between the two workloads (TQE requests are heavier).
• Persist it as a new model property in ModelProperties and pass it to the TQE run instead of the hardcoded value.
• Default to the current value (10) when unset, so existing resources behave unchanged.
• Same UI behaviour as the pre-translation field: when no resource-specific value is set, the field shows the effective default; saving a value equal to the default stores "not set".