Drop Branchless clamp logic due to performance regression

The java internal implementation of Math#min / max is faster, due to the `@IntrinsicCandidate`, enables the JIT compiler to use native method for more efficient performance.

**Only for** reference: (Added asterisk marks to avoid people who can't see)

| Benchmark            | Mode  | Cnt | Score     | Error    | Units  |
|----------------------|-------|-----|-----------|----------|--------|
| clampInline   | thrpt | 5   | 37046.833 | ±454.413 | ops/ms |
| clampInternal | thrpt | 5   | 37355.122 | ±362.992 | ops/ms |