Feature Request: discoverable and configurable execution environment capacity

未指定任何文件、测试或具体入口点。首先跟踪托管会话的生命周期，以及相关 issue #1223 中讨论的提供商和资源配置。完成的标准是满足验收标准：可发现的容量、受支持的更大环境机制，以及有明确文档记录且彼此区分的容量错误。

Problem

Sessions that run in a hosted environment receive a fixed allocation of CPU, memory, and disk. That allocation is not visible before the session starts and is not configurable. For a large repository, the first indication that capacity is insufficient is a failure partway through dependency installation, a build, or a test run — after the setup time has already been spent.

Large repositories, native application builds, and multi-platform validation all need substantially more capacity than a small allocation provides. There is also no documented pattern for delegating validation that the hosted environment cannot perform to external CI and consuming the result.

What is missing

• Effective CPU, memory, disk, operating system, and timeout values are not reported to the caller.
• There is no way to select a larger or differently configured environment where one is available.
• Caching behavior across sessions is undocumented.
• There is no documented pattern for delegating unsupported platform validation to external CI.

Proposed behavior

• The session reports its effective environment — operating system, CPU count, memory, available disk, timeout, and cache configuration — before work begins.
• Repository or organization owners can select from approved environment profiles.
• Documented guidance exists for delegating validation that cannot run in the hosted environment to external CI, and for the session to consume those results.
• Capacity-related failures are identified as such, rather than surfacing as generic build or test errors.

Example scenario

A caller starts a session against a large repository. The session reports its capacity up front, so the caller knows immediately whether a full build is viable or whether validation should be delegated to external CI.

Acceptance criteria

• Effective environment capacity is programmatically discoverable.
• At least one mechanism exists to request a larger environment where supported.
• Capacity exhaustion produces a distinct, documented error.

Related

• #1223 — sandbox support, which overlaps on lifecycle, providers, and resource configuration.