Optimize tail-padding reuse [NFC]

[kripken]

One some ABIs, if a base class ends with padding, derived classes right after it
can reuse that padding, placing their data there. This PR optimizes the key
Expression class that way.

Size changes on linux64 (common classes):

class	old size	new size
LocalGet	24	16
LocalSet	32	24
Load	64	48
Store	80	64
Unary	32	24
Binary	40	32
StructGet	40	24
StructSet	48	32

On 5 real-world binaries I tested, this reduced peak RAM usage by
3%, 4%, 6%, 9%, 9%.

[aheejin]

Nice idea!

Gemini suggested we can make these further improvements. I haven't checked them by myself though. (Also not necessarily in this PR)

1. Call: Move bool isReturn to the top (64 -> 56 bytes)
2. CallIndirect: Move bool isReturn to the top (80 -> 72 bytes)
3. CallRef: Move bool isReturn to the top (64 -> 56 bytes)
4. I31Get: Move bool signed_ to the top (32 -> 24 bytes)
5. AtomicRMW: Move MemoryOrder order to the top (64 -> 56 bytes)
6. AtomicCmpxchg: Move MemoryOrder order to the top (64 -> 56 bytes)
7. StructRMW: Move MemoryOrder order to the top (48 -> 40 bytes)
8. StructCmpxchg: Move MemoryOrder order to the top (48 -> 40 bytes)
9. ArrayRMW: Move MemoryOrder order to the top (48 -> 40 bytes)
10. ArrayCmpxchg: Move MemoryOrder order to the top (56 -> 48 bytes)
11. SIMDReplace: Move uint8_t index to the top (40 -> 32 bytes)