Perf: replace std::unordered_map with boost::unordered_flat_map/set

[sbillig]

Description

Replaces all uses of std::unordered_map/set with boost unordered_flat_map/set, and replaces std::unordered_multimap with boost::unordered_multimap. This provides a nice speedup on the purplebench suite locally (though results were noisy on my laptop so take this with a grain of salt). I'd appreciate a benchmark run on the dedicated machine.

The only real risk here is flat_map's reference (in)stability across inserts. I don't see any cases where this is currently a problem. There may be some instances where unordered_map is better than flat_map, eg where the value type is large, or in cases with a lot of updates. I ran the BalancerV3Vault benchmark case, flat_map vs unordered_map for these instances locally and flat_map was never worse, but this could use more testing:

(these are cases where the value is non-trivial (eg std::set), and/or there's potentially a lot of mutation)
CommonSubexpressionEliminator.h:172
ControlFlowSideEffectsCollector.h:130
CommonSubexpressionEliminator.h:66
DataFlowAnalyzer.h:184
SSACFGBuilder.h:159

I'm sure using flat_map in more places was on everyone's radar; thought it was worthwhile running this experiment. Happy to roll back individual uses or whatever.

contract	no via-IR	via-IR
WETH9	-2.69%	-6.95%
ChainlinkLINKLinkToken	-4.73%	-3.71%
MakerDAODSSJug	-4.53%	-4.98%
MakerDAODSSPot	-4.50%	-4.37%
ENSRegistryWithFallback	-0.76%	-4.15%
CompoundIIICometRewards	-2.90%	-3.57%
MakerDAODai	-4.55%	-5.31%
WrappedBitcoinWBTC	-3.25%	-4.64%
UniswapV4V4Quoter	-3.56%	-3.26%
TetherToken	-3.42%	-3.62%
LidoWstETH	-5.30%	-2.93%
ENSETHRegistrarController	-4.29%	-3.05%
MakerDAODSSVat	-4.82%	-7.90%
AaveV3VariableDebtTokenInstance	-2.93%	-4.29%
UniswapPermit2	-5.00%	-3.87%
OETH	-4.29%	-2.89%
BoredApeYachtClub	-3.16%	-2.18%
AaveV3ATokenInstance	-2.63%	-3.90%
RocketDepositPool	-4.27%	-2.69%
Poap	-3.89%	-3.92%
FraxFinanceFRAXStablecoin	-3.38%	-5.27%
EigenLayerEigenPodManager	-0.26%	-3.48%
SafeGnosisSafe	-4.46%	-3.79%
CoWProtocolGPv2Settlement	-4.25%	-4.10%
EigenLayerStrategyManager	-2.29%	-2.47%
FiatTokenV2_2	-6.71%	-4.66%
LidoAccountingOracle	-6.59%	-2.86%
UniswapV2UniswapV2Router02	-4.95%	-3.30%
Morpho	-6.42%	-3.36%
SparkProtocolPool	-3.13%	-2.20%
AccessControlledOCR2Aggregator	-7.40%	-4.37%
PoolManager	-2.22%	-4.79%
UniswapV4PositionManager	-1.07%	-3.82%
CompoundIIICometWithExtendedAssetList	-2.86%	-1.30%
EulerSwap	-3.30%	-3.58%
UniswapV3SwapRouter02	-4.28%	-1.99%
AaveV3PoolConfiguratorInstance	-8.55%	-3.48%
EtherfiLiquidityPool	-4.12%	-3.53%
AaveV3PoolInstance	-1.23%	-5.45%
ENSNameWrapper	-6.82%	-4.21%
UniswapV3NonfungiblePositionManager	-4.19%	-4.55%
AaveV4HubInstance	-6.78%	-2.82%
LidoWithdrawalQueueERC721	-6.57%	-1.39%
UniswapV3Pool	-2.65%	-1.08%
BalancerV2Vault	-3.38%	-2.16%
BalancerV3Router	-3.48%	-0.85%
BalancerV3Vault	-6.15%	-2.46%
Total	-3.88%	-3.29%

Checklist

• I have read the contributing guidelines and the review checklist
• I have personally reviewed, understood, and tested every change in this PR

AI Disclosure

This is just a find-and-replace PR, and AI did it. AI also ran benchmarks for me.