Improve writeFloat function for precision

Refactor writeFloat function to improve float encoding logic.

Author: mstdokumaci

src/msgpack.zig

@@ -1819,12 +1819,11 @@ pub fn PackWithLimits(
 
         /// write float
         fn writeFloat(self: Self, val: f64) !void {
-            const tmp_val = if (val < 0) 0 - val else val;
-            const min_f32 = std.math.floatMin(f32);
-            const max_f32 = std.math.floatMax(f32);
-
-            if (tmp_val >= min_f32 and tmp_val <= max_f32) {
-                try self.writeF32(@floatCast(val));
+            // A value should only be encoded as f32 if it can be 
+            // represented exactly without loss of precision.
+            const val_f32: f32 = @floatCast(val);
+            if (val == @as(f64, val_f32)) {
+                try self.writeF32(val_f32);
             } else {
                 try self.writeF64(val);
             }