feat: integrate vAMM into market-sim

[cdummett]

closes #671

PR adds the plumbing for integrating AMMs into market-sim and adds some AMM agents into the fuzzing for quick wins.

Examples:

Run a benchmark scenario with a static AMM (no base price updates).

python -m vega_sim.scenario.amm.run -m static -s 1000 -o -p

AMM Scenario:

An AMM scenario has been added which modifies the benchmark scenario by adding a single AMM agent and a mean-reverting Ornstein-Uhlenbeck price process.

Plots show the basic mechanisms of the AMM agent behaving as expected.

• short for prices > base price
• long for prices < base price
• loses money as the price diverges from the base price
• gains money as the price converges on the base price

Monte-Carlo Testing:

The AMM scenario was run multiple times with the following OH parameters:

• sigmas = [0.2, 0.4, 0.6, 0.8, 1.0]
• thetas = [0.000, 0.005, 0.010, 0.015, 0.020]

Plots show expected behavior.

• slower reversion to the mean (base price) increases AMMs risk
• greater volatility increase AMMs returns "ceiling" but increases risk of loses

[daniel1302]

I think it should be good to merge. I have tested it in live environment(stagnet1) and all is working well. Here is a bots PR: vegaprotocol/research-bots#16