If a consumer registered after a cp goes to flushing state,
the on_switchover_cp cb will not be called for this consumer.
In this CP, the ctx for this consumer is nullptr as the consumer
never participant in the cp.

Previous code calling cp_flush for every consumer, leaving the duty
of properly handle the nullptr returned by cp->context(svc_id) to
consumer. However, none of the existing consumer handled the case.

As a result, we hit an occurance that Index generate a CP sololy,
but before the cp fully flushed, other consumer registered and be
called into cp_flush(), the replication service, doesnt properly
handled the nullptr like below, `get_repl_dev_ctx` was called with this_ptr
is null, it is dangerous as invalid memory get accessed.

This change is a breaking change for consumer like HO so bump up the version.
HomeObject participant the CP as CLIENT, current implementation of HO always
returns nullptr for `on_switchover_cp` which will result the CLIENT be excluded
from cp_flush after this commit merged.

callstack:
```
homestore::ReplSvcCPContext::get_repl_dev_ctx (this=0x0, dev=0x56010ab52b00) at /home/ubuntu/HomeStore/src/lib/replication/service/raft_repl_service.cpp:521
0x0000560106d58f1e in homestore::RaftReplServiceCPHandler::cp_flush (this=<optimized out>, cp=0x56010a467940) at /home/ubuntu/HomeStore/src/lib/replication/service/raft_repl_service.cpp:549
```

code:

```
 auto cp_ctx = s_cast< ReplSvcCPContext* >(cp->context(cp_consumer_t::REPLICATION_SVC));
 ...
 auto dev_ctx = cp_ctx->get_repl_dev_ctx(repl_dev.get());
```

Signed-off-by: Xiaoxi Chen <[email protected]>

…p buffer