chore(release): prepare for 0.4.8

[github-actions]

Prepares v0.4.8, bumped by cargo release patch.

The version and the changelog are the whole diff. Merging this tags
v0.4.8 and starts the release, which builds the binaries and
plays the strength match against the previous release.

Close it without merging to call the release off, nothing has been
tagged or published yet.

[github-actions]

Speed against 55ac7f86

Measured on AMD EPYC 7763 64-Core Processor.
Both sides built with rustc 1.98.1 (48a229cea 2026-09-01).

Both sides built and run on this runner in this job, the way
scripts/speed.sh measures a perf commit. Each round runs
both sides on a layout of its own, the same compiled code with
its code and data shuffled and moved, so the 95% interval carries
where the code landed as well as the run. The verdict holds the
interval against a 1% threshold. The default layout, the one a
release ships, is measured after as a diagnostic. The node
counts are the search's: they move when the search does, and a
speed change leaves them alone.

round     base nps  candidate nps  change
    1      4728575        4656696   -1.5%
    2      4621117        4655391   +0.7%
    3      4616575        4720090   +2.2%
    4      4715434        4558374   -3.3%
    5      4625907        4643777   +0.4%
    6      4630061        4676088   +1.0%
    7      4523837        4586447   +1.4%
    8      4678732        4599810   -1.7%
    9      4517021        4514737   -0.1%
   10      4772522        4641765   -2.7%
   11      4708917        4688129   -0.4%
   12      4645578        4574255   -1.5%
   13      4645752        4678449   +0.7%
   14      4254620        4486820   +5.5%
   15      4112240        4128429   +0.4%
   16      4460994        4197248   -5.9%
   17      4145627        4157994   +0.3%
   18      4465866        4313578   -3.4%
   19      4276462        4205882   -1.7%
   20      4545167        4417096   -2.8%
   21      4595604        4659161   +1.4%
   22      4620372        4618884   -0.0%
   23      4710393        4683367   -0.6%
   24      4566465        4661200   +2.1%
   25      4659485        4687816   +0.6%
   26      4487205        4599076   +2.5%
   27      4650184        4656045   +0.1%
   28      4577737        4655613   +1.7%
   29      4581470        4558496   -0.5%
   30      4676300        4630003   -1.0%
   31      4714756        4687020   -0.6%
   32      4699751        4605755   -2.0%
   33      4646371        4322802   -7.0%
   34      4241974        4081022   -3.8%
   35      4237449        4167436   -1.7%
   36      4187009        4173001   -0.3%
   37      4605573        4297189   -6.7%
   38      4237326        4246446   +0.2%
   39      4258136        4248142   -0.2%
   40      4283281        4106723   -4.1%

             nodes    time  median nps  faster half
base       5965973  1.30 s     4600588      4668618
candidate  5965973  1.30 s     4592762      4655207
change                           -0.2%        -0.3%

paired change -0.6%, 95% interval -1.5% to +0.1%
diagnostic, on the default layout alone +0.2%, 95% interval -0.7% to +0.8%, 13 rounds

not resolved: the interval reaches past ±1.0% without clearing it, so
more rounds are needed to say either way

Speed: -0.6% (bench nps, 95% interval -1.5% to +0.1%, 40 interleaved rounds over shuffled layouts vs 55ac7f86)

The instructions each side's bench executed, counted under
cachegrind. The count repeats to within a few hundred
instructions, so a small change here is a real one, but it
prices instructions only: cache misses, mispredicted branches
and where the code lands are the speed's to show.

instructions against 55ac7f86, one cachegrind run a side, bench

            instructions    nodes  per node
base       9,380,787,246  5965973    1572.4
candidate  9,380,787,418  5965973    1572.4
change            +0.00%             +0.00%

within ±0.7%, as far as edits not made for speed have moved the count