feat: add tracing instrumentation for command execution

[joshrotenberg]

Summary

Adds rich tracing spans and events to every Docker command execution path, behind an optional (default-on) tracing feature.

• Top-level docker.command info span on CommandExecutor::execute_command with fields command, args_count, platform, runtime, timeout_secs. Because every DockerCommand::execute impl (regular, compose, buildx, swarm, manifest, generic) funnels through this method, all paths are instrumented automatically.
• info! on success with duration_ms, exit_code, stdout_len, stderr_len.
• warn! on non-zero exit / spawn failure with exit_code, duration_ms, error, plus a 512-byte UTF-8-safe truncated stderr_snippet.
• Streaming variants wrapped in a docker.stream span (mode = handler or channel); each stdout/stderr line emits debug! so noisy builds can be filtered at the subscriber level.
• New tracing cargo feature enabled by default. cargo build --no-default-features still compiles -- existing use tracing::*; sites go through a small internal tracing_compat shim that re-exports the real macros when the feature is on and expands to no-ops otherwise. #[instrument(...)] attributes are gated with cfg_attr.
• README gains a Tracing section documenting span names, fields, and RUST_LOG recipes.

Closes #241

Test plan

• cargo fmt --all -- --check
• cargo clippy --all-targets --all-features -- -D warnings
• cargo clippy --all-targets --no-default-features -- -D warnings
• cargo test --lib --all-features (768 passed)
• cargo test --lib --no-default-features (519 passed)
• cargo test --doc --all-features (408 passed)
• New tests/tracing_instrumentation.rs exercises span name, args_count field, and warn-on-failure event using tracing-test (gated behind the tracing feature, no Docker daemon required)
• Pre-existing integration tests that require a running Docker daemon were not re-run -- they fail in this sandbox due to no local daemon, not due to changes in this PR