Revert batched hashing of String/ByteArray

In 03b129b the implementation of String.hash and ByteArray.hash was
changed to hash bytes in chunks of 2/4/8 bytes at a time. While this
speeds up the hashing process, it also degrades the quality of the
resulting hash. This in turn can result in a significant increase in
hash collisions.

Refer to
#836 (comment) for
more details.

This reverts commit 03b129b.

Author: yorickpeterse

std/src/std/bytes.inko

@@ -38,38 +38,13 @@ fn pub inline check_slice_range(start: Int, end: Int, size: Int) {
 
 fn hash_bytes[T: Bytes, H: mut + Hasher](bytes: ref T, hasher: mut H) {
   let mut len = bytes.size
-  let mut cur = bytes.pointer
+  let mut start = bytes.pointer
+  let mut i = 0
 
-  len.hash(hasher)
+  hasher.write(len)
 
-  while len > 8 {
-    let val = (cur as Pointer[UInt64]).0 as Int
-
-    val.hash(hasher)
-    cur = ptr.add(cur, 8)
-    len -= 8
-  }
-
-  while len > 4 {
-    let val = (cur as Pointer[UInt32]).0 as Int
-
-    val.hash(hasher)
-    cur = ptr.add(cur, 4)
-    len -= 4
-  }
-
-  while len > 2 {
-    let val = (cur as Pointer[UInt16]).0 as Int
-
-    val.hash(hasher)
-    cur = ptr.add(cur, 2)
-    len -= 2
-  }
-
-  while len > 0 {
-    (cur.0 as Int).hash(hasher)
-    cur = ptr.add(cur, 1)
-    len -= 1
+  while i < len {
+    hasher.write(ptr.add(start, i := i.wrapping_add(1)).0 as Int)
   }
 }
 

std/test/std/test_string.inko

@@ -339,6 +339,7 @@ fn pub tests(t: mut Tests) {
     }
 
     t.not_equal(hash('hello'), hash('world'))
+    t.not_equal(hash(('ab', 'c')), hash(('a', 'bc')))
   })
 
   t.test('String.+', fn (t) {