fix: clear message goes to all bus instances + refactor: to optimize bus sync

[gkachru]

Since the channel name is the same, clear is called on all L1 instances of any cache stack with a sync bus. Clearing one namespace will clear the local caches on remote instances for ALL namespaces. Issues related to this are masked, since this only impacts the L1 cache and the L2 cache of other namespaces are still preserved.

However, since this will cause unnecessary processing of messages and clearing of L1 caches, I have added a namespace to the channel name of the bus to overcome this.

[gkachru]

Also implemented a namespace cache in CacheStack to avoid unnecessary allocations. Recommend using emittery as the default event emitter to make bus message handling async.

[gkachru]

Added a refactor of the bus related code, which is now optimised when namespaces are used:

• Single bus per named cache (reduces Redis connections when large number of namespaces are used)
• Support for sending namespaces on bus messages (including using binary encoding) and using it to access the appropriate local cache instance
• Caching of namespaced stacks to reduce memory allocations
• Added and modified existing test cases to cover more bus synchronization scenarios

All of the above is necessary if you use hundreds of namespaces and don't want to get hit by heavy memory usage and a high volume of bus messages that stall the main thread.

[Julien-R44]

Looks perfect to me. Many thanks for these quality PRs 👌