Requested feature

Docling allows to generate details timing for all the conversion stages. This feature is activated as

from docling.datamodel.settings import settings
settings.debug.profile_pipeline_timings = True

The result will be stored in

# ...
conv_res = converter.convert(...)
conv_res.timings  # < -- here

We would like to add a CLI option docling --profile INPUT which generates a summary of the profiling, e.g. for each stage "count", "total", "mean", "median", "min", "max", "0.1 percentile", "0.9 percentile".