Autoinject code? Coverage/performance/others simplification

**Describe the feature you'd like**
`byexample` allows to run arbitrary code at some points during the execution. These *hooks points* are described in the `Concern` interface.

One use case for these *concerns/hooks* is to enable coverage: the coverage system is initialized at the begin of the execution and at the end it generates a dump with the stats.

An example of this is the [coverage.py](https://github.com/byexamples/byexample/blob/91fb6b3915e9e88b44a047e8c030b89d05890db6/test/coverage.py) of `byexample`.

Other cases could be:
- enabling measure of performance for the whole execution or per example
- enabling debug logs

These are the *orthogonal* things for what `Concern` was designed for.

But creating a `Concern subclass`, implement some of its methods in Python it may be to complex, specially for people coming from other languages.

It could be much nicer just say:

```ruby
byexample -l ruby --enable-ruby-coverage
```

But that is perhaps too much for `byexample`.

A tradeoff could be:

```ruby
byexample -l ruby --inject "ruby:begin:" --inject "ruby:end:"
```

More cryptic, sure, but more general too: `byexample` could inject a *ruby file* at the begin and at the end of the execution.

`byexample` doesn't need to know about coverage/performance/debug/etc and the user doesn't need to know about Python and the `Concern` interface.