fix(PositionPlanner): cap liquidity by per-tick headroom instead of a plan-wide budget

[EperezOk]

Summary

• PositionPlanner.resolve bounded Uniswap v4's per-tick maxLiquidityPerTick with a single plan-wide budget decremented after each position, conservatively treating the whole plan as if every position shared one tick boundary (which skips positions v4 would accept).
• This change tracks liquidityGross per boundary tick and caps each position (and the implicit full-range fallback) to the headroom remaining at its own boundaries — matching v4's actual per-tick rule.

Background

In v4, maxLiquidityPerTick is enforced per tick against that tick's liquidityGross, and liquidityGross only
accrues at a position's two boundary ticks (tickLower/tickUpper). Two positions therefore compete for the cap only when they share an exact boundary; positions on distinct ticks are independent and may each hold up to
maxLiquidityPerTick.

The previous guard collapsed this into one plan-wide budget (Σ(all plan liquidity) ≤ maxLiquidityPerTick). It guaranteed no mint could overflow a tick on a fresh pool, but at the cost of dropping valid positions.

Change

_headroom(...) sums the liquidityGross already contributed by created positions at each of a candidate's two
boundaries and returns maxLiquidityPerTick − max(grossLower, grossUpper). Each candidate (and the fallback) is capped to that value before its liquidity is quoted.

Properties preserved:

• No overflow / no revert on a fresh pool. Capping each position to its boundary headroom keeps every boundary's gross ≤ maxLiquidityPerTick (by induction, so maxLiquidityPerTick − maxGross can't underflow). Enforcing the per-tick gross bound is also what keeps the active-liquidity uint128 accumulator from overflowing.
• Graceful degradation unchanged. Insufficient headroom shrinks or skips a position (rolling its allocation into the fallback); it never reverts.
• Budget invariant unchanged. A position's quoted amounts never exceed its budget, so the checked subtraction in resolve still can't underflow.

Behavior change

Positions on independent boundaries now each receive their own per-tick headroom instead of competing for one budget, so plans that previously had positions silently dropped will now mint them. The existing
test_resolve_clampsPositionsAboveMaxLiquidityPerTick characterization test was updated to reflect this (a clamped explicit position now coexists with the full-range fallback rather than starving it).

Gas

Measured A/B of resolve (old vs new) on identical inputs where neither version clamps, isolating the per-tick scan cost:

Positions	old	new	overhead
1	23,989	25,010	+1,021
2	29,452	31,601	+2,149
5	59,574	67,699	+8,125
10 (max)	110,064	136,797	+26,733

End-to-end, the standard 2-position migration snapshot moves 803,742 → 806,053 (+2,311 gas, < 0.3%); the worst case (10 positions) adds ~+26.7k to an ~830k-gas migration (~3%). The LBPStrategy_E2E_Test.json gas snapshot is updated accordingly.