Improve Docker daemon diagnostics for local flowctl runs

[RitwijParmar]

Summary

Fixes #2969 by detecting Docker daemon/connectivity failures from docker stderr even when the Docker CLI exits successfully.

Docker Desktop on macOS can return status 0 while writing the real daemon failure to stderr. Before this change, flowctl preview / flowctl catalog test could continue into docker inspect parsing and surface a misleading error like:

failed to parse `docker inspect` output: []

This now returns an actionable message telling the user to start Docker Desktop or their configured container runtime, with a note about DOCKER_CLI=podman.

What changed

• Adds docker_cmd_output in both runtime paths:
  • crates/runtime/src/container.rs
  • crates/runtime-next/src/container.rs
• Classifies daemon-unavailable stderr before checking exit status, covering the macOS case where Docker exits 0 but writes Cannot connect to the Docker daemon....
• Preserves existing behavior for ordinary non-daemon Docker failures.
• Adds snapshot regression tests in both crates for:
  • daemon unavailable with successful exit status and [] stdout
  • non-daemon Docker command failure passthrough
• Documents the Docker/container-runtime prerequisite for local flowctl preview and flowctl catalog test workflows in crates/flowctl/README.md.

Validation

Passed locally:

cargo fmt --check
cargo test -p runtime docker_cmd_ --lib
cargo test -p runtime-next docker_cmd_ --lib