A number of performance metrics have regressed as of switching to the v5 toolchain in SERVER-101790; some of these regressions appear to be a result of v5 compilers making different inlining decisions than those in v4. We should use forced inlining in the short-term to buy back these regressions, while leaving ourselves the option in the future to remove these forced inlining directives once we're able to turn on PGO and other optimizations that should result in more consistent, stable, and optimal binary layouts.

Add a macro that resolves to MONGO_COMPILER_ALWAYS_INLINE and whose name indicates clearly that the purpose of the forced inlining is to resolve v5 regressions. Then, update the changes in SERVER-104864 so that the changes to forcibly inline functions in that ticket use the new macro.