When using the RLOG DB Backend with Mria, replicant nodes do not
generate `mnesia down` events. Therefore, cleanup procedures that
some modules do when a node goes down do not work for replicants.
However, replicant do generate `node down` events, so that may be a
safer way to handle cleanup to take into account that type of node.
* feat(gateway): add gateway application
* chore(gateway): add normalize confs function
* refactor: move emqx-stomp to emqx-gateway subdir
* chore(gateway): fix some bad function defination
* chore(gateway): rename type to gwid
* chore(gw-stomp): upgrade the implementation to suppport gateway instance
* feat(gw-stomp): add reconnect mechanism
* refactor(stomp): upgrade connection&channel module to latest
* refactor(stomp): more details for handle_in/out
* refactor(stomp): get it up and running
* chore(gw): load some modules by default
* refactor: upgrade the emqx-gateway schema module
* test(stomp): fix testcases for stomp gateway
* chore(gw): remove needless lines
* chore(gateway): correct a lot of specs
* chore(gw): add a draft for metrics
* chore(gw): add metrics process
* fix(gw): fix cm process monitor
* test(gw): add test cases for gateway-regitry
* feat(gw): add metrics/cli for gateway
* fix(gw): fix xref errors
* chore(gw): pretty gateway metrics print format
* chore(gw-stomp): generate clientid by default
* chore(gw): more reliable
* chore(gw): rename gwid -> type
* chore(gw): impl the update logic
* chore(gw): some format improvement
* chore(gw): adapts the hocon configs
* fix(gw): fix xref errors
* test(gw): update configurations for tests
* chore(gw): ignore diaylzer warnings
* fix(gw): fix bad function call
* chore(gw): remove needless comments
Zaiming (Stone) Shi