Debug traces longer than max Grpc message

[ferranbt]

Rationale

Why should this feature exist?

With the bor debug command we get 'heap', 'cpu' and 'trace' traces from the running client. Both 'cpu' and 'trace' run for n seconds to get the trace, n is an flag in the cli. However, if the trace runs for more than 4 seconds, the output data is too big to be transferred over a single Grpc message.

Implementation

Do you have ideas regarding the implementation of this feature?

Use something like https://github.com/mitchellh/go-grpc-net-conn to create a stream connection over the Grpc connection to send arbitrary amount of data and pipe that stream into a file.