tools: add support for validation scripts in test files

[fjl]

This adds a facility for defining 'validation scripts' in test files. Example file:

>> {"jsonrpc":"2.0","id":1,"method":"eth_gasPrice"}
<< {"jsonrpc":"2.0","id":1,"result":"0x1047435"}
--
if (BigInt(messages[1].response.result) <= 0) {
    throw new Error("gasprice too low");
}

The validation script runs after message exchanges are completed, and can access all data in the test,
as well as the server's responses. Scripts are executed during test filling (to verify the script is syntactically correct), and will also be run in the hive runner with the actual server responses.

To make this work, I have refactored the handling of .io files a bit and there is now a Go package
for working with them. This package will be used in hive to load the files and run the scripts.

[MysticRyuujin]

A couple of things maybe worth fixing?

1. Script timeout. RunScript runs goja with no interrupt or deadline. --sim.timelimit defaults to zero. We might want to consider a deadline via vm.Interrupt or something?

2. Loader tests. #1613 deletes hive's two loader tests (the 24 MiB single-line case and the comment/speconly parsing case) along with the old parser, but they did not move into iofile_test.go. The large-line test exists because fixtures with big receipts once broke the default scanner buffer. Should we port those two into the iofile package so that coverage does not go away?

[fjl]

Thanks for feedback. I will check it about the loader test. I enlarged the token buffer a bit to make it work.

[fjl]

Regarding execution timeout, it's a good point. I will add this.

[lightclient]

This is awesome! So ideally this will also solve ethereum/hive#1588 which it looks like you already commented on.

In that scenario, I think the way it will work is that we'll check the full schema for the tracer and if that passes we know the schema is correct. Then it should just be a matter of verifying that the type of tracer matches the expected, i.e. if we're trying to test opcode tracer we use the JS script to just check that the response was that tracer type and not another. We've already validated the schema so we know it is at least properly formed.

Does that seem correct @MysticRyuujin ?

[fjl]

I have added a jsonschema.validate(schema, value) function to the scripting environment, and the script now has access to the OpenRPC spec through the openrpc global variable. So it is now theoretically possible to validate results against a specific object schema. Ideally, we can have a script like:

let schema = openrpc.components.schemas.OpcodeBlockTransactionTrace;
jsonschema.validate(schema, messages[1].response.result);

This doesn't work at the moment because we operate on the dereferenced spec and it doesn't contain the components.schemas anymore.

[fjl]

Ahh, I realize that making this work also requires a different API for the jsonschema module in JS. In order to resolve references in the spec, the validation function needs access to the whole spec object. So it will look more like this:

let schema = "#/components/schemas/OpcodeBlockTransactionTrace";
jsonschema.validate(openrpc, schema, messages[1].response.result);