All documented benchmarks and timings should mention the Mill version

[lefou]

Mill changes and improves with any version and numbers benchmarked with a specific version that looked good today may mis-reflect the abilities of later releases. We already have versioned documenation pages, but since the benchmarks and many comparison pages (vs. Maven, Gradle, sbt) are all hand written, we have no way to ensure up-to-date correct numbers. Hence, we should explicitly state the tool versions used to measure.

Since I expect much better numbers once we have a native Mill client, It would also help to identify the benchmarks and comparison pages which need an overhaul.