exercises(binary-search): example: improve (#254)

Changes:

- Make `buffer` no longer an optional. This means we now never return
  `SearchError.NullBuffer`, but there was no test case for that.

- Make the implementation more standard, and avoid overflow when
  calculating `mid`.

I intend to later make this exercise return an optional, not an error
union.

Refs: #229

exercises/practice/binary-search/.meta/example.zig

@@ -1,32 +1,23 @@
 pub const SearchError = error{
     EmptyBuffer,
-    NullBuffer,
     ValueAbsent,
 };
 
-pub fn binarySearch(comptime T: type, target: T, buffer: ?[]const T) SearchError!usize {
-    if (buffer == null) {
-        return SearchError.NullBuffer;
-    }
-    if (buffer.?.len == 0) {
-        return SearchError.EmptyBuffer;
-    }
-    var i: usize = 0;
-    var j = buffer.?.len - 1;
-    while (i < j) {
-        const mid = (i + j) >> 1;
-        if (buffer.?[mid] < target) {
-            i = mid + 1;
-        } else if (buffer.?[mid] > target) {
-            j = mid - 1;
-        } else {
+pub fn binarySearch(comptime T: type, target: T, buffer: []const T) SearchError!usize {
+    if (buffer.len == 0) return SearchError.EmptyBuffer;
+
+    var left: usize = 0;
+    var right = buffer.len;
+
+    while (left < right) {
+        const mid = left + (right - left) / 2; // Avoid overflow.
+        if (buffer[mid] == target) {
             return mid;
+        } else if (buffer[mid] < target) {
+            left = mid + 1;
+        } else {
+            right = mid;
         }
     }
-    // Unroll when i == j, otherwise the function logic above might cause
-    // integer overflow to occur at the edges of usize.
-    if (buffer.?[i] == target) {
-        return j;
-    }
     return SearchError.ValueAbsent;
 }