feature: replace the old parsing strategy and move towards a tree based one

[eliax1996]

The scope of this pr is to replace the old parsing strategy and move towards a tree based type verification strategy.

From this commit it's possible to refer to a type using a partial path and not only a fully specified path. If we declare for e.g. the following schema:

syntax = "proto3";

package my.awesome.customer.plan.entity.v1beta1;

message CustomerPlan {
  string plan_name = 1;
}

When importing it it's sufficient to use a part of the full specified path (that in this case would be my.awesome.customer.plan.entity.v1beta1.CustomerPlan). As e.g. it would be enough to use entity.v1beta1.CustomerPlan or v1beta1.CustomerPlan, so the following declaration would be accepted:

syntax = "proto3";

package my.awesome.customer.plan.entity.v1beta1;
import "my/awesome/customer/plan/entity/v1beta1/entity.proto";

message CustomerPlanEvent {
  message Created {
    entity.v1beta1.CustomerPlan plan = 1;
  }
}

[eliax1996]

This pr it's experimental, even if it's passing all the tests stuff can be still moved around and a final judgement should be given from @jjaakola-aiven before going forward. Please try as much valid (and invalid) schemas as possible and provide feedbacks if something it's fishy.
A sample test it's the test_partial_path_in_protobuf, feel free to start from here and craft new ones.

[aiven-anton]

I left some minor nits.

[eliax1996]

Can I have another round of review @jjaakola-aiven and @aiven-anton?
I still have to get rid of the # pylint: disable=W0212 and split the huge test in 2/3 tests (because they are testing different aspects related but independent about the parser).

Except that wdyt? I've tried to transform the TypeTree into an immutable object but the descending visit would become much harder to read, I think that in that case the penalty of having a mutable object is ok if the parsing algorithm it's easier. We can also create an internal object called _TypeTree with a List in it and a type visible externally where we replace the List with a Sequence, would it be a good enough tradeoff to keep mutability inside and immutability outside @aiven-anton?