[FIXED] Fixes for filestore FSS state

[neilalexander]

This PR contains some bug-fixes for the filestore FSS state.

First, if the filter provided is _EMPTY_ then we need to overwrite it to > for the stree functions to work, but we failed to set wc when doing so. This could affect the decision on whether to linear-scan or not. We will now set wc too in this case.

Second, if the FSS only contains a single subject then we may not correctly set isAll if the filter contained a wildcard, as Find does not work with wildcards and therefore wouldn't match the single subject. Updated to use Match instead, so that it correctly uses wildcards.

Third, replaces Match(>) with Iter in enforceMsgPerSubjectLimit, not because it is behaviourally different, but because we need to walk the entire stree anyway and Iter saves CPU cycles by skipping the match step.

Fourth, updates the bounds checking on various functions that use Find so that the logic matches the Match equivalents. This also matters in particular in SubjectsState which could fail altogether when the filter was a subject literal rather than a wildcard and the psim was out-of-date/needing update.

Fifth, in NumPending and in filteredPendingLocked when matching the FSS state, if we find a partial, don't process any further matches after that. This bug was introduced in #5559 as there used to be a break there, whereas now the Match that replaces it can't be interrupted in the same way.

Signed-off-by: Neil Twigg neil@nats.io

[derekcollison]

LGTM - Small comments but really good changes.

[derekcollison]

LGTM