Change ReadPackage to print a warning if specified file cannot be read (e.g. because it is missing)

[fingolfin]

Resolves #5745 but unfortunately can't be merged as-is because it turns out several packages read files that don't exist... But I didn't want to loose these changes, so here they are as a PR anyway. (In fact I modified this PR to only print a warning instead of just erroring out

How to proceed? Some options:

1. we update all affected packages to not do this anymore, then once they are all updated in the package distro, we merge this but with the error turned into a warning
2. same as 1, but we keep the warning as a warning (possibly using InfoWarning ?)
3. some clever alternative I can't think of right now...

[ChrisJefferson]

I think a warning is best for now. I'd be happy to merge the warning earlier (maybe 90 days? I dunno, just making up a random deadline), if packages haven't updated by then they'll start getting a harmless warning, and we still get plenty of time before the next release in case there are other issues (for example in packages which aren't distributed, if their devs or users use master)

[fingolfin]

Suggested change

	Print( "ReadPackage could not read <", pkgname, ">/", relpath, "\n" );
	# TODO: turn this into an `Error` in a future GAP version
	Print( "ReadPackage could not read <", pkgname, ">/", relpath, "\n" );

[fingolfin]

Also, should we use Info(InfoWarning, 1, "...") instead? Or Print but with some prefix like #I ?

[fingolfin]

I think this is basically good to go now, given that (I think) no package in the distro triggers the warning anymore. To be clear: it still is just a warning, IMHO we should wait with turning this into an error for at least a full release cycle.

Would be good to get some final feedback on this.

[ThomasBreuer]

The changed documentation promises an error, which we do not get (yet). Is this o.k.?

From the viewpoint of Oscar, it would be good to be able to suppress the message, and since Oscar already sets the level of all GAP info classes (in particular InfoWarning) to 0 by default, using Info(InfoWarning, 1, "...") would fit.

In practice, currently this would not make a difference, and eventually the warning should indeed be turned into an error, in order to make ReadPackage behave like Read. In this sense, I have no strong opinion about Info vs. Print.

[ChrisJefferson]

I'd probably do InfoWarning, just because that feels natural, and means it can be turned off easily if it really upsets someone (Oscar, Sage) for some reason.

[fingolfin]

Switched to InfoWarning and adjusted the documentation.

[ThomasBreuer]

Thanks.