[APM][Otel] Refactor the normalize function into per data structure serialization logic

[jennypavlova]

In this issue we are adding a function to normalize the fields - It is because the elasticsearch fields API is returning a different data structure than the _source we used before.
In the first version, we ended up with a generic normalization function which we want to refactor here. We want to have different mappers for every type used to map only the fields needed in the type:

Example TransactionRaw

Response example:

with _source: {......trace: {id: 123},......}
with fields: {.....'trace.id': ['123'],.....} // not something we can change with the query

normalize so we have something like (it could be a separate function per type replacing the current normalization) :
Transaction mapper->
{
processor: [event.processor][0];
timestamp: [timestamp.us][0];
trace: { id: [trace.id][0] };
....APMBaseDoc fields,
....
}

This will improve the structure and make sure that we have only the fields needed in the type and the corresponding values.

AC:

• All normalization should be replaced with a mapper function to a corresponding type (transaction, span, error, etc)
• We should test that it works the same for apm server using apm indices and for Otel data using otel indices

[elasticmachine]

Pinging @elastic/obs-ux-infra_services-team (Team:obs-ux-infra_services)

[miloszmarcinkowski]

Linked PR: jennypavlova#6