Add compare_release fast path (#799)

Looking at the profile for tf-models-nightly after #789,
`compare_release` is the single biggest item. Adding a fast path, we
avoid paying the cost for padding releases with 0s when they are the
same length, resulting in a 16% for this pathological case. Note that
this mainly happens because tf-models-nightly is almost all large dev
releases that hit the slow path.

**Before**


![image](0d2b4553-da69-4cdb-966b-0894a6dd5d94)

**After**


![image](6d484808-9d16-408d-823e-a12d321802a5)

```
$ hyperfine --warmup 1 --runs 3 "target/profiling/main pip-compile -q scripts/requirements/tf-models-nightly.txt"
 "target/profiling/puffin pip-compile -q scripts/requirements/tf-models-nightly.txt"
Benchmark 1: target/profiling/main pip-compile -q scripts/requirements/tf-models-nightly.txt
  Time (mean ± σ):     11.963 s ±  0.225 s    [User: 11.478 s, System: 0.451 s]
  Range (min … max):   11.747 s … 12.196 s    3 runs

Benchmark 2: target/profiling/puffin pip-compile -q scripts/requirements/tf-models-nightly.txt
  Time (mean ± σ):     10.317 s ±  0.720 s    [User: 9.885 s, System: 0.404 s]
  Range (min … max):    9.501 s … 10.860 s    3 runs

Summary
  target/profiling/puffin pip-compile -q scripts/requirements/tf-models-nightly.txt ran
    1.16 ± 0.08 times faster than target/profiling/main pip-compile -q scripts/requirements/tf-models-nightly.txt
```

crates/pep440-rs/src/version.rs

@@ -2092,6 +2092,9 @@ impl<'source> FromPyObject<'source> for Version {
 /// Compare the release parts of two versions, e.g. `4.3.1` > `4.2`, `1.1.0` ==
 /// `1.1` and `1.16` < `1.19`
 pub(crate) fn compare_release(this: &[u64], other: &[u64]) -> Ordering {
+    if this.len() == other.len() {
+        return this.cmp(other);
+    }
     // "When comparing release segments with different numbers of components, the shorter segment
     // is padded out with additional zeros as necessary"
     for (this, other) in this.iter().chain(std::iter::repeat(&0)).zip(