Agent refactor (#1487)

This PR is the 3rd PR in a series of PRs for agent0 to support multiple
pools.

The agent classes in agent0 is reworked in this PR. Here, the main
changes include `init_agent` being a function of the chain as opposed to
the pool. Additionally, many functions now take a `pool` argument to
specify what pool the agent is making a trade on. There are helper
functions for setting an active pool to avoid passing this argument in
various functions, by calling it either in `init_agent`, or by calling
`agent.set_active(pool=...)`.

Additionally, the policy is now untied from the agent, such that an
agent can hot-swap policies on the fly by calling the
`agent.set_active(policy=...)` function.

- `init_agent` is now a function of a `LocalChain` or `Chain` object.
- This function now takes an optional `pool` argument for setting an
initial active pool.
- Many functions now require an active pool to be set, or an explicit
pool to be passed in as an argument.
- `PolicyAgent` has been deprecated in favor of using `LocalAccount`.
- Moving many logging and bookkeeping configs from `Hyperdrive` and
`LocalHyperdrive` to `Chain` and `LocalChain` respectively.
- No longer keeping track of wallet deltas, we get the wallet wrt a pool
via `get_wallet()`.
- Adding helper functions for `get_long()`, `get_short()`, `get_lp()`,
and `get_withdrawal_shares()`.
- `LocalHyperdrive` now has a flag for deploying on constructor, or
attach to existing. This is necessary for the forking workflow.
- Added tests for unit and system fuzzing.

examples/interactive_hyperdrive_forking_example.py

@@ -9,7 +9,7 @@
 
 from fixedpointmath import FixedPoint
 
-from agent0 import Hyperdrive, LocalChain, PolicyZoo
+from agent0 import LocalChain, LocalHyperdrive, PolicyZoo
 from agent0.core.base.make_key import make_private_key
 
 # %%
@@ -26,26 +26,22 @@
 # Launch a local anvil chain forked from the rpc uri.
 chain = LocalChain(fork_uri=rpc_uri, fork_block_number=fork_block_number)
 
-hyperdrive_address = Hyperdrive.get_hyperdrive_addresses_from_registry(chain, registry_address)["sdai_14_day"]
+hyperdrive_address = LocalHyperdrive.get_hyperdrive_addresses_from_registry(chain, registry_address)["sdai_14_day"]
 
-# Note that we use Hyperdrive here instead of LocalHyperdrive,
-# as LocalHyperdrive deploys a new pool, whereas we want to connect to an existing pool
-# on the forked local chain.
-# TODO this prevents us from using data tools provided by LocalHyperdrive, ideally we can
-# load a LocalHyperdrive from an Hyperdrive object that connects to an existing pool and populates
-# the database. This is blocked by needing an archive node, the fix here would be to
-# (1) use event data instead, and (2) build historical data from event data.
-hyperdrive_config = Hyperdrive.Config()
-hyperdrive_pool = Hyperdrive(chain, hyperdrive_address, hyperdrive_config)
+# Note that we pass in deploy=False and pass in an existing hyperdrive_address, as we
+# want to connect to the existing pool and not deploy a new one.
+hyperdrive_config = LocalHyperdrive.Config()
+hyperdrive_pool = LocalHyperdrive(chain, hyperdrive_config, deploy=False, hyperdrive_address=hyperdrive_address)
 
 # %%
 
 # Launch a new agent
 private_key = make_private_key()
 
 # Init from private key and attach policy
-hyperdrive_agent0 = hyperdrive_pool.init_agent(
+hyperdrive_agent0 = chain.init_agent(
     private_key=private_key,
+    pool=hyperdrive_pool,
     policy=PolicyZoo.random,
     # The configuration for the underlying policy
     policy_config=PolicyZoo.random.Config(rng_seed=123),