This is the same code I used in the `v0.4-stable` backport.  I didn't
use it in the main branch because the minimum ruby version is 3.1.  But
that breaks when JRuby and TruffleRuby try to `require "net/imap"`.

Fixes #452.

Bumps [step-security/harden-runner](https://github.com/step-security/harden-runner) from 2.11.1 to 2.12.0.
- [Release notes](https://github.com/step-security/harden-runner/releases)
- [Commits](step-security/harden-runner@c6295a6...0634a26)

---
updated-dependencies:
- dependency-name: step-security/harden-runner
  dependency-version: 2.12.0
  dependency-type: direct:production
  update-type: version-update:semver-minor
...

Signed-off-by: dependabot[bot] <[email protected]>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

The bug was that exclusive ranges ending in zero would be converted to
end on `-1`, which would be interpretted as the last value in the range.

This maybe isn't actually documented very well (or at all...) but most
SequenceSet transform methods return a frozen result when +self+ is
frozen and a mutable result when +self+ is mutable.  Except +limit+
which always returns a frozen result.  And (before this commit) +slice+,
which inconsistently returned with matching frozen status when the
result wasn't empty, but always returned a frozen set when the result
_was_ empty.

Adding these tests exposed a much more significant bug:
`SequenceSet#xor` mutates the reciever.

`set ^ other` and `set.xor other` are supposed to be safe transforms.
But, unfortunately, they modified the receiver if it wasn't frozen, and
crashed when it was!

The fix is trivial: convert `self` to `dup`.

The goal is for `#[]` (aliased as `#slice`) to behave similarly to
`Array#[]`/`Array#slice`.  When `Array#slice` has a length or range that
extends beyond the end of the array, they simply return everything
up to the end.

When `count` is given to either of these methods, a new `SequenceSet` is
returned, containing only the last `count` numbers.  An empty
SequenceSet is returned when `self` is empty.

The goal is for these to behave similarly to `Array#min`/`#max` when
they are given a count.

✨ Add `SequenceSet#min(count)` and `#max(count)`

These are just sugar over `set - (..num)` and `set - (num..)`.  They are
included because they can sometimes make the code much easier to read
than doing set algebra with `-` or `&`.

I'm not sure how this escaped testing before, but yikes!

I almost accidentally committed a big bug for `#xor`.  Yikes!  I decided
to take that opportunity to simply add some randomized tests on all of
the set operators, based on set identities that should always hold true.

These can also be used for microbenchmarks and profiling of SequenceSet.

* Links back to "Ordered and Normalized Sets".
* Small updates to `*` description.
* Small updates to `min`, `max`, `minmax` summary.
* Added more "related" links.
* Added more "Set identities" sections to set operators.

✅ Test `SequenceSet#xor` and fuzz test all set operations

📚 Various `SequenceSet` rdoc improvements