feat: execution trace gantt chart

[viktorasm]

When debugging highly parallelized task executions, it's not immediately obvious where the bottlenecks are.

This PR allows dumping timings of the execution as Gantt chart (Mermaid syntax). Here's a sample of of a real-word invocation:

Having this makes it so much easier to understand, how well things parallelize, are tasks unnecessarily blocked, what are the major bottlenecks, etc.

Usage:

• if user provides --execution-trace-output flag, then task invocation writes contents of the gantt chart to it.

[trulede]

This could be interesting a general technique.

For instance, a tracer could dump each tasks variables as tasks execute, making it much easier to debug Taskfiles.

[trulede]

Here perhaps an interface, and then implementation for the Mermaid tracer.

type Tracer interface {
    Start(task)
    Stop()
    ...
}

[trulede]

I think here, the same pattern as other executors should be used (above calls).

[trulede]

Might it be possible to consider a more generic operation:

# Select one tracer or the other, with default options.
task default --tracer=mermaid
task default --tracer=csv

# Select a tracer and also provide some options, without adding additional CLI parameters.
task default --tracer="mermaid;filename=foo.out"

[trulede]

PR #2617 introduces some logging features which could be a more generalised alternative to what this PR is doing. Then it would be a matter of processing the log into the format you need with a simple script.