Using `duct` as a library (in a separate project)

_Originally posted by @CodyCBakerPhD in https://github.com/con/duct/issues/327#issuecomment-3618198391_

> So far I haven't got any real-world need for it-- unless youve got one?

I would love to create a benchmarking library that leverages this :) (mentally reserving `con/vent` for this purpose)

Our work on [NWB Benchmarks](https://github.com/NeurodataWithoutBorders/nwb_benchmarks/blob/main/src/nwb_benchmarks/benchmarks/time_download.py#L34) relied on hacking a version of Airspeed Velocity, which set out to do _way_ too much as a package that essentially wrapped `timeit` with an info-capturing layer, rather like `duct`, and as a result was a pain to work with (even though we eventually got something out of it)

Bonus: such benchmarks could serve as the 'batch' call Yarik has previously requested

The proposal is for a separate benchmarking library that leverages duct, and it references NWB Benchmarks' src/nwb_benchmarks/benchmarks/time_download.py and Airspeed Velocity. Start by reading that example and the prior batch-call request mentioned in the issue. The issue does not define implementation scope or clear completion criteria.

Written by the indexing model from the issue text.