Improve performance for test generation and execution

[MatheusFranco99]

Overview

The purpose of this PR is to optimize the performance of test generation and execution with the purpose of improving work quality on future changes.

The ssv module is the one that takes the most time. After running a CPU profiling test, we noticed that the NewTestingKeyManager function consumes 80% of the total time, which is explained by the fact that it's called multiple times and takes 26 milliseconds to run (according to benchmark tests).

Nonetheless, the NewTestingKeyManager function returns a testingKeyManager object that is only used to validate and sign data but does not change state throughout a test execution. Thus, we propose adding a singleton pattern for this object.

Metrics

To illustrate the number of times the NewTestingKeyManager function is called, take a look at the following table

Note: the number after the dash (/) is the number of different instances created, which depends on a slashableDataRoots parameter.

Module	Test generation	Test execution
QBFT	368/1	390/1
SSV	17058/1	20470/2
Types	48/1	48/1

Performance comparison

With the singleton pattern, if the appropriate object already has been created, the function takes 5 microseconds, rather than 26 milliseconds.

All time values shown below are in seconds (or in minutes:seconds if it has ":" ).

For test generation, we got the following performance results.

	QBFT	SSV	Types
Old	15	7:36	3
New	5	20	3

For test execution, we got the following performance results.

	QBFT	SSV	Types
Old	16	1:22	3
New	4	5	2

[GalRogozinski]

Great!

[GalRogozinski]

you can consider using MuHash or something else which is faster

[y0sher]

fnv is easy fast and simple for tasks like this. built in the go stdlib

[MatusKysel]

better approach would be getHash before locking, so we keep it locked as short as possible

[y0sher]

generally I would say global state is a bad idea, but since this is a test generation code and you are protecting it properly, its seems ok to me.