test(fork): exercise TailTwins tail-match gate off parity

[aaronjmars]

What

Adds an off-parity fork test for TailTwins. Every ForkDeploy fixture initializes at SQRT_1_1 (2^96), whose low byte is 0, so TailTwins.requiredTail is a constant 0 and the "your size's last byte must match the price's, which shifts every swap" mechanic is never actually exercised at a nonzero tail.

test_TailTwins_tailBindsOffParity inits a Base pool off parity (sqrtPrice low byte 0x37) and asserts:

1. requiredTail != 0 (the 1:1 blind spot is avoided),
2. a size whose low byte is a half-ring away reverts TailMismatch,
3. a tail-matching size clears the gate and moves the price, so the gate is proven to read live slot0, not a frozen constant.

Verification

forge build clean. All 3 GateFix tests pass on a live Base fork:

TT requiredTail @init 55
TT off-tail swap reverted TailMismatch
TT matching-tail swap OK; requiredTail now 214

Behavioral coverage only; no source or deployment change. Closes the last sibling whose gate was vacuous at parity (HeavierHand + CapGate were already covered off-parity in the same file).