Feat/per service concurrency limit #1381

[Chapoly1305]

This pull request introduces per-service concurrency limits for GPT-based translation services, allowing users to control the maximum number of simultaneous requests for each service. The implementation includes a concurrency manager, updates to service definitions and settings dialogs, and localization for the new setting in English, Chinese, and Italian.

Concurrency control core implementation:

• Added a ServiceConcurrencyManager class and a Semaphore utility to manage and enforce per-service concurrency limits, ensuring that no more than the specified number of requests are processed simultaneously for each service. [1] [2]
• Integrated the concurrency manager into the main addon state and the translation task runner, so all translation tasks for GPT services are now subject to concurrency control. [1] [2] [3] [4]

Service and settings updates:

• Extended the TranslateService interface to support an optional concurrencyLimit property, and updated the GPT service factory to read and apply this limit from user preferences. [1] [2]
• Added a numeric "Concurrency Limit" setting to the GPT and AzureGPT service dialogs, including support for custom GPT services, with a placeholder indicating that 0 means unlimited. [1] [2]
• Registered new concurrency limit preferences for each GPT service in prefs.js. [1] [2]

Localization and documentation:

• Updated English, Chinese, and Italian locale files to include labels and placeholders for the new concurrency limit setting. [1] [2] [3]
• Documented the new concurrency limit feature in the README.md, explaining its purpose and usage.

[windingwind]

we probably want to reuse the logic of resource://zotero/concurrentCaller.mjs from Zotero itself?