when yielding data from a resource, mark given value to be normalized

**Goal**
We want to mark certain string values to be normalized according to naming convention used later in the pipeline. Compare such feature to `dlt.with_table` where we mark whole dictionary to be routed to table with given type.

**Motivation**
Quite often some lineage metadata is loaded with the data itself. For example in case of `pipedrive` we load a table with definitions of user defined fields. Those fields are present in `dlt` schema and their names are normalized like any other column names.

**Tasks**
1. * [ ] add `normalize` function in `dlt.mark` module. that will wrap a `str` in a string like object that will be serialized in `typed_dump` as a special type (like for example `Decimal` or `datetime`)
2. * [ ] when deserialized in normalizer, a `normalize_identifier` method from the current schema context will be applied to the value

Start in the dlt.mark module and trace how typed_dump serializes special values such as Decimal or datetime. Then follow deserialization in the normalizer to the current schema context and its normalize_identifier method. Done means marked strings survive serialization and are normalized according to that schema context.

Written by the indexing model from the issue text.