This patch introduces the _generation_ value for the event cache lock:
every time the lock is acquired by a new holder, the generation is
incremented by 1. This is fundamental to know if the lock has been
poisoned or not.

We are about to introduce a lock which checks if it's been poisoned. To
not break the current API semantics, `lock` is renamed `lock_unchecked`
if poisoning isn't check/doesn't raise an error.

…kingStore`.

This patch implements `BackingStore` for all `Arc<T>` where `T:
BackingStore`.

This patch implements `LockableEventCacheStore:is_poisoned` to know
whether the lock has been poisoned or not.

This patch implements `EventCacheStoreLock::lock`, which is similar
to `lock_unchecked` but it returns a `Result<EventCacheStoreLockGuard,
EventCacheStoreLockPoisonError<…>>` to handle the _poisoned_ case.

This patch adds a small `logged_in_base_client_with_store_config`
function similarly to `logged_in_base_client`, but… with… a store
config!

This patch adds tests for `EventCacheStoreLock::lock` and
`lock_unchecked`.