Idea: Allow user control over state

[flash1293]

Currently state is managed transparently by the user by storing it in the cache and automatically using it if present.
The only control that's granted to the user is to ignore the current state.

There are some cases where this doesn't work properly:

• No way / no desire to retain the cache (e.g. because no local files can be kept, like in a Github action)
• When using streaming via get_records

In these cases it would be nice to still allow a user to do incremental syncs by managing the state themselves. This means there is a method to get the current state as an opaque string, which can be stored whatever and be passed in again:

state_str = cache.dump_state()
# persist this string somewhere
# ...
# in another run
source.read(cache, state=state_str)

For get_records:

my_stream = source.get_records("my_stream")
for r in my_stream:
  # do something

# can also be called if the iterator is not exhausted yet - it keeps track of the latest state messages
state_str = my_stream.dump_state()

# in another run
my_incremental_stream = source.get_records("my_stream", state=state_str)