Add printing of LoaderAllocator to MethodTable in SOS

[janvorli]

This change adds printing of LoaderAllocator to MethodTable dump in SOS
for collectible types.
It also fixes one issue I had in the object members iterator for LoaderAllocator for the case when an object has no other object references than the LoaderAllocator via MethodTable.

[janvorli]

@dotnet/dnceng There seems to be something wrong with the CentOS7.1 x64 Checked Innerloop Build and Test leg. I have two PRs and in both of them this leg has been running for about 10 hours and viewing the leg console output shows just this:

02:07:46 GitHub pull request #20255 of commit 3efcaf5ed0dee1df2b708e5837105d80e341344c, no merge conflicts.
02:07:47 Setting status of 3efcaf5ed0dee1df2b708e5837105d80e341344c to PENDING with url https://ci.dot.net/job/dotnet_coreclr/job/master/job/x64_checked_centos7.1_innerloop_flow_prtest/7655/ and message: 'Started.'
02:07:47 Using context: CentOS7.1 x64 Checked Innerloop Build and Test

[MattGal]

@janvorli taking a look now.

[MattGal]

@janvorli it would seem that machine got into some kind of hung state (was idling and had failed a ton of other jobs). I deleted that VM and the new flow job seems to be going OK. I'll synch with @mmitche when I can to see if there was a less destructive way to repair this, but I think that will work. The Windows ARM runs are still going to be mega slow until we resolve the challenges from #20204 (specifically: answer where CoreCLR jobs will get native ARM smarty from)

[janvorli]

@MattGal thank you!