Clarify that lowerCamelCase field names MUST be used for OTLP/JSON

[tigrannajaryan]

Resolves #2795

This is a breaking change for OTLP/JSON and is allowed because OTLP/JSON is not yet Stable.

[tigrannajaryan]

I discussed this with @yurishkuro.

He has a very good argument that we control most of OTLP producers (the SDKs and Collector), while we don't control most of OTLP consumers (e.g. the backends).
In a tradeoff where we need choose between simplicity for producers and and simplicity for consumers we prefer to choose simplicity of consumers at the cost of more complication for producers. We prefer to wear the burden ourselves instead of burdening the consumers.
I agree with this design choice. Because of this I support the following:

• Producers must only produce lowerCamelCase JSON keys. They must not produce original field names as JSON key. Doing so is trivial, since it is the default behavior of Protobuf JSON marshalers and only requires producers to be careful and avoid choosing a non-default marshalling option. It is also trivial to comply with for producers that don't use Protobuf JSON marshalers.
• Consumer don't need to worry about accepting original field names. They should only accept lowerCamelCase field names.
  The OTLP/JSON specification will unconditionally call out that lowerCamelCase names are required for JSON keys.

I will update this PR to reflect this thinking and will reset the existing approvals.

[tigrannajaryan]

Let's give this another 24 hours and then merge if there are no objections.

[tigrannajaryan]

Has enough approvals. No objections for several days. Merging.