[processor/spanprocessor] Implement context-aware/structured span processing capabilities

[lopes-felipe]

Component(s)

processor/span

Is your feature request related to a problem? Please describe.

Currently, this processor's span transformation capabilities are based on regex and/or attribute names without any contextual awareness of their origin (e.g., SQL Query, Redis Command, Web Request, etc.).

This makes it hard to apply more complex processing logic, such as removing comments or collapsing parameters for IN-clauses (ref), as well as extracting any common attribute defined by the semantic conventions (ref).

Describe the solution you'd like

The ability to apply transformation logic to the spans, based on a configuration. Ex:

span:
  name:
    sql:
      comments: <keep|remove>
      alias: <keep|remove> # whether SQL aliases ("AS") should be truncated
      space_between_parentheses: <keep|remove> # IN ( ? ) | IN (?)
      ...
  attributes:
    sql:
      table_name: <insert|upsert> # db.collection.name (users; dbo.products)
      operation: <insert|upsert> # db.operation.name (SELECT; INSERT; UPDATE; DELETE; CREATE; mystoredproc)
      query_summary: <insert|update> # db.query.summary (SELECT orders; INSERT shipping_details; UPDATE users)
      ...

Describe alternatives you've considered

No response

Additional context

This feature is based on this Datadog-specific obfuscation process PR. However, the idea here is to provide something similar but much more flexible, complete, and vendor-agnostic.

I can definitely take this, if approved :)

[github-actions]

Pinging code owners:

• processor/span: @boostchicken

See Adding Labels via Comments if you do not have permissions to add labels yourself.