config: allow four byte-size config to be specified using human-readable units ("100 GiB")

[kennytm]

What problem does this PR solve?

Several config specifies a large byte size, which is not very readable (and some DBAs are still unfamiliar with the underscore-as-number-separator TOML feature).

What is changed and how it works?

Using docker/go-units (already used by Dumpling and PD), we support parsing human-readable byte size for the following config:

• tikv-importer.region-split-size (100_663_296 → '96 MiB')
• mydumper.read-block-size (65536 → '64 KiB')
• mydumper.batch-size (107_374_182_400 → '100 GiB')
• mydumper.max-region-size (268_435_456 → '256 MiB')

Byte size can be specified using integers (65536 or 65_536, backward-compatible with existing config), floating point (6.5536e+4), or a string involving byte units ('64k', '64 K', '64KiB', '64 kb', '0.0625 MB', all equivalent).

(This PR is submitted mainly because we expect that disk-quota is also going to need some huge numbers in a custom config, similar to TiKV's ReadableSize struct. Interestingly no such thing exists in TiDB except in dealing with INSPECTION_RULES.)

Check List

Tests

• Unit test

Side effects

Related changes

• Need to update the documentation
• Need to be included in the release note

Release notes

• The 4 configuration specified above can now accept human-readable format in the form "2.5 GiB".

[overvenus]

LGTM