Improve replication document error when user_ctx or auth headers are not parsed correctly

Adding something like
```
"user_ctx": null
```
to the replication document results in the state `failed` which is correct but the error reason is: `"{error,{case_clause,null}}"` which is not the best.

Same goes for using a `null` value for a header in a source/target endpoint object.

Start by reproducing replication with a null user_ctx and with a null header in a source or target endpoint object. Trace the replication document and endpoint error handling to find where the case_clause error is produced; done means these inputs still fail but report a clear, useful error reason.

Written by the indexing model from the issue text.