Profile and time deploy operations

[wowi42]

Summary

Provide built-in profiling and timing for pyinfra deploys so users can see how long each operation takes and identify bottlenecks.

Motivation

Wrapping the whole run with time gives a single overall number, which is not enough to understand where time is spent. Per-operation timing would help users:

• Compare deploy speed across instance types and connectors.
• Identify slow operations or facts.
• Diagnose regressions after upgrades.

Proposed scope

• Time both phases of every operation: prepare/generate and execute.
• Time fact gathering per host and per fact.
• Print a final summary at the end of the run, similar to Finished, took 11 min 35 s, plus a breakdown of the slowest operations and facts.
• Optionally include timestamps on each -vvv log line so users can correlate timings with individual commands.
• Expose timing data programmatically via the API or as a structured output (JSON) so it can be consumed by external tooling.

References

Consolidates and supersedes:

• Profiling deploys #469 - Profiling deploys
• Feature request: time measurement / operation profiling #640 - Feature request: time measurement / operation profiling