Fix Capy positioning + interop documentation (value proposition, execution invariant, escape hatch)

[sgerbino]

Summary

The review showed reviewers reaching incompatible conclusions about what Capy
is, because the docs don't state its value proposition, its core execution
invariant, or its interop escape hatch. The authors agreed these are real
problems.

Detail

Three related doc gaps:

• Value proposition / "What Capy is and is not": Reviewers disagreed on
  whether Capy is a universal coroutine-interop library or the non-socket layer
  of a specific I/O library. A "What Capy is / is not" statement should be
  folded into the docs.
• Core execution invariant, stated up front: State that "a coroutine will
  always be resumed by the same Executor object which was used to launch it,"
  and why that guarantee matters.
• Foreign-awaitable escape hatch: There is a documented way to co_await a
  plain/foreign awaitable, but it is undocumented; a reviewer hit the
  type-safety compile check, concluded no escape hatch existed, and this fed a
  reject vote.

Acceptance criteria

• Front-page "What Capy is / is not" statement of the value proposition and scope.
• The same-executor-resumption invariant is documented up front with rationale.
• The foreign-awaitable escape hatch is documented with an example.

References

Overlaps existing open issue #341 ("Docs: Highlight the IoAwaitable dimension of
the library"). Consider folding into / linking with #341.

Raised by Rainer Deyke, Christian Mazakas, Peter Dimov; Vinnie Falco
acknowledged the gaps. Source: Boost ML announcement thread + "[Capy] Review
Note".