bench(spann): add a Criterion harness for the distance kernels

[ohdearquant]

What

Adds a Criterion benchmark harness for ruvector-spann's distance module. The
crate has distance kernels in src/distance.rs: l2_squared is on the index
hot path (src/kmeans.rs, src/index.rs), and cosine_distance is a public
kernel with no production call site today, covered by the same harness. Until
now the crate had no benches/
directory, no [[bench]] target, and no criterion dev-dependency — so a
performance change to this crate cannot be measured inside the crate itself.

No behaviour change. The only Cargo.toml additions are under
[dev-dependencies] (criterion, rand, both at the workspace-pinned
versions) plus the [[bench]] declaration.

Design notes

• One Criterion group per public distance function, parameterized over
  dimensions 128 / 384 / 768 / 1536.
• Operands come from a seeded StdRng (gen_range(-1.0..1.0)), not constant
  fills, so runs are reproducible while still exercising data-dependent
  behaviour. Several existing benches elsewhere in the workspace use
  constant-fill vectors, which say nothing about data dependence; this harness
  deliberately does not inherit that.
• black_box on both inputs and outputs.

Verification

• cargo bench -p ruvector-spann --bench distance -- --test: all 8 cases
  print Success.
• cargo fmt --all -- --check: clean.
• cargo clippy -p ruvector-spann --all-targets: no Clippy diagnostic names
  the new bench file. Pre-existing, unrelated output remains: Cargo's
  workspace profile-ignored warnings and the ruvector-attention bench/bin
  double-target warning, both untouched here.

Why now

#763 proposes an optional distance backend for this crate and currently has no
way to carry before/after numbers. This harness is the missing instrument, kept
as its own PR so the instrument and the change it would measure are reviewed
separately.

---

Note on this PR's CI. Tests (core-and-rest) is red on every branch in this repository,
including main: the job is cancelled at its 240-minute cap while still compiling and never
reaches the test phase. #786 restores the exclusion list that the shard's packages: value loses
to a shell comment, #784 unblocks the ruvector-filter test target that the compiler cannot
finish, and #787 fixes a deadlock waiting behind both. That failure is not caused by this branch.