Revert some Semigroups-related downrankings of Matrix methods

[wilfwilson]

Three downrankings of Matrix methods were added in advance of GAP 4.10, along with one for BaseDomain, in order to make sure that GAP 4.10 did not break the Semigroups package. This was supposed to be temporary, and #2758 exists to revert this change.

In fact, only one of these downrankings is required to keep Semigroups happy. This commit reverts the others. Once it is merged, I will update #2758 accordingly.

At some point (separately from this PR) I will try to see about removing the remaining downranking of a Matrix method.

[coveralls]

Coverage increased (+0.0005%) to 84.443% when pulling 47d259f on remove-downranking into fcd1f51 on master.

[fingolfin]

Looks good to me. I did not verify how this affects Semigroups or anything else, though -- i.e., I am trusting @wilfwilson on this (and am happy to do so).

I guess this could/should be backported to stable-4.11, right?

[wilfwilson]

There is an additional FIXME that I didn't consider - I'll check it out shortly.

[wilfwilson]

I've also reverted the BaseDomain downranking that also existed (thanks @PaulaHaehndel for pointing it out to me). Doing so doesn't seem to affect the Semigroups package, so I don't think there can be any harm in removing it.

@fingolfin As for backporting, I don't mind. These downrankings didn't seem to be having any consequences, so I don't know whether backporting would make any noticeable difference to anything. Please add the backport tag if you think it's appropriate.

[wilfwilson]

Just to clarify, I ran the GAP testinstall, teststandard, testbugfix test files with this branch of GAP and with the latest released version of the Semigroups package loaded, and I also ran the Semigroups package tests in this version of GAP. Everything passed as normal.