.m.rule.tombstone and .m.rule.room.server_acl pushrules are in the wrong order

[richvdh]

Link to problem area:

https://spec.matrix.org/v1.5/client-server-api/#default-override-rules

Issue

.m.rule.tombstone and .m.rule.room.server_acl are listed as happening before .m.rule.roomnotif; in practice (see synapse implementation, previously in python) they take place after .m.rule.roomnotif.

In practice, it's unlikely that a given event will match .m.rule.roomnotif as well as one of those two rules, so it doesn't matter that much, but it's an unnecessary source of confusion.

The problem appears to have arisen from matrix-org/matrix-spec-proposals#2020, which inexplicably inserted .m.rule.tombstone into the middle of the list, in contrast to the implementation (linked above) and MSC1930 (which didn't explicitly mention an order, so "the end" was the most plausible interpretation).

This was then compounded when .m.rule.room.server_acl was added by MSC3786, saying "immediately after .m.rule.tombstone". The interpretation of that obviously depended on where .m.rule.tombstone appeared in the list.

I believe the spec has always been wrong here, and .m.rule.tombstone and .m.rule.room.server_acl should be moved after .m.rule.roomnotif.

(Note also that Synapse implements an .m.rule.reaction, from MSC2677).