Extend benchmark to make it more essential

[abelsiqueira]

The benchmark should help decide if a PR is to be accepted or not. So we should improve the script to test more.

• Goal: Make sure that there are no regressions
• Test multiple solvers in multiple problems
• No need to test in all problems since we are not comparing the solvers, only the JSOSuite side
• Change script to run whenever src/ files are modified. Edit: this is already the case