Marginal Filmic Chroma speedup with new techniques

yeah I didn't realize you could just cast an array of bytes into an
array of multi-byte primitives my bad. Code smaller and like 30% faster.

bsz-filmic-chroma/filmic-chroma.rs

@@ -1,4 +1,3 @@
-use std::convert::TryInto;
 use std::os::raw;
 
 #[no_mangle]
@@ -9,19 +8,12 @@ pub extern "C" fn filmic_chroma(
     bytes: *mut raw::c_char,
     len: usize) {
     let pixels = unsafe {
-        std::slice::from_raw_parts_mut(bytes.cast::<u8>(), len)
+        std::slice::from_raw_parts_mut(bytes.cast::<f64>(), len/8)
     };
-    for chunk in pixels.chunks_exact_mut(32) {
-        let l = f64::from_le_bytes(chunk[0..8].try_into().expect("bytefail"));
-        let c = f64::from_le_bytes(chunk[8..16].try_into().expect("bytefail2"));
-
-        let c = match invert {
-            false => c * (offset - l / scale),
-            true => c * (offset - (100.0 - l) / scale)
-        }.to_le_bytes();
-
-        for x in 0..8 {
-            chunk[x+8] = c[x];
+    for pixel in pixels.chunks_exact_mut(4) {
+        pixel[1] = match invert {
+            false => pixel[1] * (offset - pixel[0] / scale),
+            true => pixel[1] * (offset - (100.0 - pixel[0]) / scale)
         };
-    }
+    };
 }