Add a new kernel header gap_all.h for use by package authors instead of compiled.h

[fingolfin]

So far, most package kernel extensions #include "compiled.h" in order
to conveniently get access to (almost) all source headers of GAP at once
(which is nice for us because it means we are free to rename headers or
move stuff around between them without breaking kernel extensions).

But the result is rather counterintuitive: if one is not familiar with
this, seeing #include "compiled.h" in code is not at all clear. Also,
the original of compiled.h was to be included by C code generated by
gac, and it contains stuff we'd rather not encourage others to use.

Thus, add a new header gap_all.h containing all #include statements
formerly found in compiled.h -- this resolves all concerns.

---

Of course it will take time for kernel extensions to switch to this new header, especially if they care about staying compatible with older GAP versions. But the sooner we add this, the sooner such a switch could happen. Hence it would be nice to backport this (any objections).

[stevelinton]

No objections. It might be worth looking for cases where packages use the other functionality from compiled.h. Some cases might be functionality that should be moved elsewhere, others might be considered errors to fix in the package.

[coveralls]

Coverage remained the same at 84.513% when pulling 130dd7a on fingolfin:mh/gap_all.h into 012238d on gap-system:master.

[DominikBernhardt]

@fingolfin It seems to me we might want to add this to the release notes so people who write kernel extensions are actually aware of this change?

[frankluebeck]

The change looks ok to me. I have checked that the Browse and EDIM packages work as usual with including gap_all.h instead of compiled.h.

[fingolfin]

@stevelinton I am pretty sure we can easily adapt any package which uses something from compiled.h that is not in gap_all.h; and even if not, there is no immediate plan to remove compiled.h, at least not on my side, so package would still keep working.

@DominikBernhardt agreed, so I added the label

@frankluebeck thanks for the test!

[stevelinton]

@fingolfin Agreed. My thinking was just that it might point us to things incompiled.h which ought to be somewhere else because they are generally useful.

[fingolfin]

Backported to stable-4.11 in commit 35ffd5f