Improve finish! behavior for Progress

[halleysfifthinc]

When finish! is called on a Progress type meter with a slow update limit and a significant number of steps left, it will max a CPU core while slowly updating the progress meter until it is complete.

MWE:

julia> p = Progress(100000000; dt=.1);

julia> next!(p); @time finish!(p)
Progress: 100%|██████████████████████| Time: 0:00:16
 14.919932 seconds (6.75 k allocations: 357.250 KiB)

This took me by surprise, and doesn't match the immediate completion behavior specified in the code for the finish! methods for ProgressThresh and ProgressUnknown.

PR:

julia> p = Progress(100000000; dt=.1);

julia> next!(p); @time finish!(p)
Progress: 100%|██████████████████████| Time: 0:00:01
  0.000068 seconds (43 allocations: 3.422 KiB)

[codecov]

Codecov Report

❗ No coverage uploaded for pull request base (master@80acddf). Click here to learn what that means.
The diff coverage is 100.00%.

@@            Coverage Diff            @@
##             master     #198   +/-   ##
=========================================
  Coverage          ?   97.27%           
=========================================
  Files             ?        1           
  Lines             ?      587           
  Branches          ?        0           
=========================================
  Hits              ?      571           
  Misses            ?       16           
  Partials          ?        0           

Impacted Files	Coverage Δ
src/ProgressMeter.jl	97.27% <100.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 80acddf...0823129. Read the comment docs.

[IanButterworth]

Indeed! This will have been especially hurt by the increase in lock-unlock speed in recent julia versions.

It seems like it should always have been this way. Thanks

[IanButterworth]

There was a Windows CI hanging bug that #195 just fixed, so I'll re-run the merge commit by closing & opening

[IanButterworth]

It seems like that didn't actually test the merge commit given CI passes on master. I'll merge