Feat/queue revamp

[Rubilmax]

This PR implements the latest specs on supply/withdraw queues. It relies on some definitions and invariants to verify execution is correct:

Definitions

• a market is enabled as long as it has a non-zero cap OR the vault supplies liquidity on it
• a market is hence disabled when it has a zero cap AND the vault doesn't supply any liquidity on it

Invariants

1. supplyQueue stores any queue of markets: if the market has a zero cap, it is skipped ; if it doesn't, it is enabled and thus supplied to
2. withdrawQueue only stores enabled markets and always stores them all

Invariant 2 is verified thanks to the way markets are enabled and the way the withdraw queue is managed:

• when a market is enabled, it is necessarily pushed to the withdraw queue
• changing the order of the withdraw queue is only performed via a permutation of enabled markets. By recursivity, withdrawQueue always contains all enabled markets

I let the reviewer prove that "changing the order of the withdraw queue is only performed via a permutation of enabled markets"!

Gas diff

• Enabling a market
  • Before: required 2 txs (submitMarket + enableMarket = 350k)
  • After: requires a single tx (submitCap + acceptCap if timelock = 122k)
• Disabling a market
  • Before: disableMarket (not tested)
  • After: submitCap(0) (122k)

[MerlinEgalite]

Partial review but nice job!