closeYieldVault can leave residual FCM position due to precision/rounding

[liobrasil]

Summary

closeYieldVault can succeed while leaving a tiny non-zero FCM position (small residual debt/collateral).

Root cause: close flow relies on UFix64 (8 decimals) plus swap quote rounding, so exact repay amounts are not always representable in one pass.

This is separate from #152.

Why this matters

• A user expects "close" to mean fully closed.
• Current behavior can leave a small residual position after close.
• Residual is usually tiny, but this breaks close invariants and creates inconsistent behavior.

Expected behavior

On close, protocol should either:

1. Fully unwind debt/collateral to zero (or a defined dust threshold), then burn; or
2. Revert and keep vault open if full unwind cannot be guaranteed.

Acceptance criteria

• After closeYieldVault, residual debt/collateral is zero (or <= explicit dust threshold).
• No successful close that burns strategy while leaving non-zero position above threshold.
• Precision edge-case tests added around UFix64 minimum unit (0.00000001).

Numerical example (what is happening)

Observed close flow (mainnet investigation):

• Before close:

  • Collateral: 0.00001000 WETH
  • Debt: 0.02038486 MOET
  • getYieldVaultBalance(): 0.00000996 WETH (less than full collateral)

• During close:

  • 0.01911391 tauUSDFv swapped to 0.02030333 MOET for repayment

• After close:

  • Residual debt: 0.00008153 MOET
  • Residual collateral: 0.00000004 WETH

Precision constraint:

• Price used: 3312.54037990 MOET/WETH
• Exact collateral needed for residual debt: ~0.00000002461253 WETH
• UFix64 step is 0.00000001, so exact amount is not representable
  • 0.00000002 underpays
  • 0.00000003 clears debt (with small overpayment)

[liobrasil]

Related tracking: #179 captures Scenario 3 close failures when FlowALP PR #163 dust workaround is absent (repro + residual values included).