Optional Chain ID in Baseline Solver Configuration

[nlordell]

@vkgnosis pointed out in #1202 that the chain ID parameter being mandatory is not practical for local deployments or when testing out on new chains.

In fact, the baseline solver only needs to know the WETH address, so this PR changes the configuration semantics to require either:

• specifying a chain-id to use the canonical WETH contract for that chain.
• specify a custom weth address to use that instead, at which point the ChainId is no longer needed

Test Plan

Adjusted test loading the example TOML configuration.

[vkgnosis]

Now we're mixing deny_unknown_fields and flatten which you mentioned before doesn't work. deny_unknown_fields has already been pretty helpful for me, but this is a small config file so doesn't really matter. More generally, I think not using flatten is nice because it makes the structure of the config file more obvious from the code. I like keeping a one-to-one correspondence of rust field to config field. With complicated serde instructions it tends to get harder to figure out how the serialized file is going to look. Again, this file is small so not super important, so up to you.

[nlordell]

Now we're mixing deny_unknown_fields and flatten which you mentioned before doesn't work

It doesn't work well. deny_unknown_fields and flatten together have a whole bunch of prickly issues:

• If you flatten a struct that also has deny_unknown_fields, then this won't work (the inner struct will see a whole bunch of fields from the parent and deny them).
• If you deny_unknown_fields on untagged enums, then this also doesn't work

However, in this particular case of deny_unknown_fields, flatten and rename on an enum with only single values, it does. I tested it out locally, but you can also see an example of this in the playground.

[nlordell]

So, I changed it to use a manual match (chain_id, weth). The reasoning is:

• As you mentioned complicated serde instructions make things harder to figure out, also the whole complex interaction between deny_unknown_fields, flatten and rename_all is not obvious (for example, if you don't use the rename_all, then serialization will not work as expected)
• You get nicer panic error messages

[MartinquaXD]

nit: Could you make the comment refer to the chain-id parameter? This makes the dependency more explicit and less error prone if somebody where to insert a new argument between chain-id and weth.