Replace criterion with gauge as the benchmark framework

[sjakobi]

This removes the circular dependency on criterion containers in the benchmarks.

One consequence of this is that we can now simply use cabal new-bench
to run the benchmarks.

[sjakobi]

For some previous discussion on using gauge, see #462 (comment) and the following comments.

gauge has developed quite a bit since. See http://hackage.haskell.org/package/gauge/changelog.

[ethercrow]

bump @treeowl

This circular dependency is really a very annoying barrier of entry to working on containers and unordered-containers.

[sjakobi]

Bump @treeowl.

At this point, gauge offers a very similar feature set as criterion:

≻ cabal run -- sequence-benchmarks -h
Up to date
Microbenchmark suite - built with gauge 0.2.4
  -I CI     --ci=CI                   Confidence interval
  -G        --no-gc                   Do not collect garbage between iterations
  -L SECS   --time-limit=SECS         Min seconds for each benchmark run, default is 5.0 in normal mode, 0.0 in quick mode
            --min-samples=COUNT       Min no. of samples for each benchmark, default is 10 in normal mode, 1 in quick mode
            --min-duration=MILLISECS  Min duration for each sample, default is MilliSeconds 30, when 0 stops after first iteration
            --include-first-iter      Do not discard the measurement of the first iteration
  -q        --quick                   Perform a quick measurement and report results without statistical analysis
            --measure-only=FILE       Just measure the benchmark and place the raw data in the given file
            --measure-with=FILE       Perform measurements in a separate process using this program.
            --resamples=COUNT         Number of boostrap resamples to perform
            --regress=RESP:PRED..     Regressions to perform
            --raw=FILE                File to write raw data to
  -o FILE   --output=FILE             File to write report to
            --csvraw=FILE             File to write CSV measurements to
            --csv=FILE                File to write CSV summary to
            --json=FILE               File to write JSON summary to
            --junit=FILE              File to write JUnit summary to
  -v LEVEL  --verbosity=LEVEL         Verbosity level
  -t FILE   --template=FILE           Template to use for report
  -n ITERS  --iters=ITERS             Run benchmarks, don't analyse
  -m MATCH  --match=MATCH             Benchmark match style: prefix (default), exact, pattern (substring), or ipattern (case insensitive)
  -l        --list                    List benchmarks
            --version                 Show version info
  -s        --small                   Set benchmark display to the minimum useful information
  -h        --help                    Show help

[sjakobi]

Bump @treeowl!

How about you just merge this and give it a try? If it turns out to be a bad idea, the change should be easily reverted!

The reduced barrier for new contributors should make this easily worth it.

[treeowl]

Okay!