feat(CtUnresolvedImport): keep all imports in noClassPath mode

[nharrand]

Related to #2927 .

[nharrand]

Ready for reviews!

[monperrus]

cool stuff, thanks! see my comments.

[nharrand]

Ok Actually I need to add a prettyprinter test. Back to WiP (The test is easy to write but it fails...)

[nharrand]

Related to #2933

[nharrand]

The current implementation should work BUT it might lead to overlapping import when pretty-printing.
An example is src/test/java/spoon/test/prettyprinter/testclasses/Rule.java.
When imports are computed,

import java.util.Arrays;
import java.util.Collections;
import java.util.Comparator;
import java.util.HashSet;
import java.util.List;
import java.util.Set;

are added (as these types are indeed used) but the import overlap with the pre existing CtUnresolvedImport import java.util.*;...

Edit: This is now solved.

[nharrand]

OK, so now (with this PR), in noClassPath mode:

1. Unresolved import in the original sources will remain when the model is pretty printed.
2. But that means that import computation will never remove them even if they are unused.

This mean, that if an import my.package.* is unresolved we don't get the nice clean up effect that we use to have.

[monperrus]

Equals is not tested https://coveralls.io/builds/23377168/source?filename=src%2Fmain%2Fjava%2Fspoon%2Fexperimental%2FCtUnresolvedImport.java#L73

This is dangerous, could you add one assertTrue and one assertFalse?

[monperrus]

Great feature for the noClasspath mode. Thanks!