Parameterize benchmarks, add throughput measurements

[dralley]

Added a throughput measurement to the macrobenchmark runs

fully_parse_document/sample_rss.xml                                                                            
                        time:   [338.77 us 343.69 us 349.75 us]
                        thrpt:  [539.25 MiB/s 548.75 MiB/s 556.72 MiB/s]       <-------------
                 change:
                        time:   [-28.438% -22.166% -15.255%] (p = 0.00 < 0.05)
                        thrpt:  [+18.001% +28.479% +39.740%]
                        Performance has improved.
Found 2 outliers among 100 measurements (2.00%)
  2 (2.00%) high severe

[codecov-commenter]

Codecov Report

Merging #418 (9cadfbe) into master (d87eff8) will increase coverage by 0.17%.
The diff coverage is 0.00%.

❗ Current head 9cadfbe differs from pull request most recent head de51d9b. Consider uploading reports for the commit de51d9b to get more accurate results

@@            Coverage Diff             @@
##           master     #418      +/-   ##
==========================================
+ Coverage   49.39%   49.57%   +0.17%     
==========================================
  Files          22       22              
  Lines       13987    13935      -52     
==========================================
- Hits         6909     6908       -1     
+ Misses       7078     7027      -51     

Flag	Coverage Δ
unittests	49.57% <0.00%> (+0.17%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
benches/macrobenches.rs	0.00% <0.00%> (ø)
src/lib.rs	12.26% <0.00%> (-0.07%)	⬇️
src/de/mod.rs	77.38% <0.00%> (+0.09%)	⬆️

---

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 d87eff8...de51d9b. Read the comment docs.

[dralley]

This does make the low-level comparative benchmarks take quite a bit longer, because obviously there's 12x as many test inputs now.

That might not be worth it, we could just select a few that give a nice representative sample. Or, the user can just filter out the ones they don't want with the standard command-line filtering.

[dralley]

Criterion makes this so easy 👌

[dralley]

xml-rs is so slow in comparison that unfortunately it makes the comparative charts almost useless. quick-xml is a tiny dot that doesn't always render :)