[ACTP] use shared procmgr client from par-control

[embeaken]

What does this PR do?

Makes par-control consume the procmgr-owned client crate introduced by #54676.

It replaces par-control's local process-manager bindings, endpoint resolution, UDS connector, Windows named-pipe connector, and ERROR_PIPE_BUSY retry loop with dd-procmgr-client. Executor-specific mTLS remains in par-control and wraps the shared platform stream.

Motivation

The process-manager daemon, its CLI, and par-control should use one implementation of platform-sensitive client transport. This prevents endpoint defaults and Windows named-pipe retry behavior from drifting, without coupling par-control to the full dd-procmgrd-lib daemon implementation.

This PR is intentionally separate from the procmgr-owned extraction so the ownership boundary remains clear: #54676 defines and dogfoods the general client API, while this PR contains only Action Platform adoption.

Describe how you validated your changes

dda env dev run -- bazel test //pkg/privateactionrunner/par-control:par-control_test
dda env dev run -- env -u PKG_CONFIG_LIBDIR cargo clippy --manifest-path pkg/privateactionrunner/par-control/Cargo.toml --all-targets -- -D warnings

The full top-of-stack par-control suite passes after rebasing every later PAR layer over this adoption.

Additional Notes

This PR sits immediately after the lifecycle layer because that is the first layer that introduces a process-manager client. Later executor transport continues using the same shared connector beneath its mTLS wrapper.

[embeaken]

Folded into #54589 by fast-forwarding the lifecycle branch to this head. GitHub therefore marked this PR merged automatically. It is no longer part of stack #54679.