CSV Transaction Processor

Run a with a test input like

cargo run -- __tests__/sample.csv > output.csv

More info on test cases in ./__tests__/readme.md

Caveats

• There's an unhandled edge case in which a dispute could make an account balance negative.
• In a distributed environment, you could put the AccountActor behind a network interface or two and pipe commands in that way.
• I didn't use Serde because it gucks up the structs when you need a high level of validation / mapping. But it would have been a perfectly valid approach.
• In the real world the account actor probably wouldn't be responsible for something as specific as "write clients to stdout".
• A lot of the error branching is panicking or logging; those would be ideally be handled.
• The errors don't currently return the transaction id, which isn't particularly useful.
• Tokio w/ features = ["full"] is probably overkill; I find it convenient to block on actors.
• We could parallelize serializing the clients before writing to stdout.