Skip tracing when CI mode disabled and manual API is used

[anmarchenko]

What does this PR do?
Couple of small but important fixes regarding CI mode configuration:

• currently when using manual API and setting settings.ci.enabled to false (when running tests locally) leads to a complete failure because @ci_recorder is nil. Now we are creating the recorder in disabled state in this case to stop tracing but preserving the behaviour.
  • In order to make it more bearable without adding null checks everywhere a new span type is provided - NullSpan which is an implementation of NullObject pattern.
• we will disable test suite level visibility always when we detect old agent version that does not support event protocol

How to test the change?
Unit tests are provided

[romainkomorn-exdatadog]

Are null spans dropped by the tracer's "writer" (whatever the component that actually sends span data to our intakes is called in the Ruby tracer)?

[anmarchenko]

Are null spans dropped by the tracer's "writer" (whatever the component that actually sends span data to our intakes is called in the Ruby tracer)?

Datadog's internal Tracer does not know anything about them, they are just created in place to avoid NullPointerException, so they are not kept in any context and are not submitted anywhere