Approach 09: Use memchr and loop instead of split (71s)

Use memchr instead of split. memchr uses SIMD for faster seraching of
needle in the haystack. The split function is slow because it contructs
the split parse everytime it is called.

Few links to read more:
* https://github.com/BurntSushi/memchr
* https://blog.burntsushi.net/bstr/#motivation-based-on-performance

Cargo.toml

@@ -15,3 +15,4 @@ rust_decimal_macros = "1.34"
 fast-float = "0.2"
 rustc-hash = { version = "1.0"}
 memmap2 = {version = "0.9.4"}
+memchr = { version = "2", default-features = false }