Merge 'Vector speedup' from Nikita Sivukhin

This PR reduces allocation for vector distance calculation and also use
[simsimd](https://github.com/ashvardanian/SimSIMD) library to execute
cosine/l2 distances for dense vectors.

Reviewed-by: Pere Diaz Bou <pere-altea@homail.com>

Closes #3802

Cargo.lock

@@ -523,10 +523,11 @@ dependencies = [
 
 [[package]]
 name = "cc"
-version = "1.2.17"
+version = "1.2.41"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "1fcb57c740ae1daf453ae85f16e37396f672b039e00d9d866e07ddb24e328e3a"
+checksum = "ac9fe6cdbb24b6ade63616c0a0688e45bb56732262c158df3c0c4bea4ca47cb7"
 dependencies = [
+ "find-msvc-tools",
  "jobserver",
  "libc",
  "shlex",
@@ -1504,6 +1505,12 @@ dependencies = [
  "windows-sys 0.59.0",
 ]
 
+[[package]]
+name = "find-msvc-tools"
+version = "0.1.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "52051878f80a721bb68ebfbc930e07b65ba72f2da88968ea5c06fd6ca3d3a127"
+
 [[package]]
 name = "findshlibs"
 version = "0.10.2"
@@ -4143,6 +4150,15 @@ dependencies = [
  "similar",
 ]
 
+[[package]]
+name = "simsimd"
+version = "6.5.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0e3f209c5a8155b8458b1a0d3a6fc9fa09d201e6086fdaae18e9e283b9274f8f"
+dependencies = [
+ "cc",
+]
+
 [[package]]
 name = "slab"
 version = "0.4.9"
@@ -4924,6 +4940,7 @@ dependencies = [
  "rustix 1.0.7",
  "ryu",
  "serde",
+ "simsimd",
  "sorted-vec",
  "strum",
  "strum_macros",