And use it in the remove processor.

Related to #123891, and also this is a follow up to #124322 and #125051 (earlier nearby PRs that were laying the groundwork for this change).

Prior to this change, we had to traverse the document tree twice in the remove processor for each field that we wanted to remove: once to check whether the field existed (in the hasField call), and then once to actually remove the field (in the removeField call). This was necessary because removeField would throw an exception if the field didn't exist, so the call had to be guarded. By adding an ignoreMissing parameter to removeField we can remove the hasField-guarding and just specify that we don't care if the field doesn't exist (well, assuming ignore_missing has been set to true on the processor itself, which it typically is in the wild).

I'm labeling this as a >refactoring since there's no user-visible change in behavior, I'm just twiddling the code a bit so that it happens to be faster. On which note, this speeds up the remove processor by 30% -- I'm seeing that it's taking 290 microseconds per document rather than 413 on main (a further note: prior to #120573 it was taking 681 microseconds per document for the same benchmark).