Add dereferenceable LLVM attribute to some GC functions

[gbaraldi]

This may allow for some more optimizations

[gbaraldi]

@nanosoldier runbenchmarks(!"scalar", vs=":master")

[gbaraldi]

I don't think that run ever went through

[gbaraldi]

@nanosoldier runbenchmarks(!"scalar", vs=":master")

[vtjnash]

@nanosoldier runbenchmarks(!"scalar", vs=":master")

[vtjnash]

@nanosoldier runbenchmarks(!"scalar", vs=":master")

[vtjnash]

@gbaraldi your group membership is private at https://github.com/orgs/JuliaLang/people/gbaraldi, so @nanosoldier will not authenticate you to run jobs (and you need a ``code fence\)

[nanosoldier]

Your benchmark job has completed - possible performance regressions were detected. A full report can be found here.