fix(span_links): ensures trace_flags is consistent with the sampling priority

[mabdinur]

What does this PR do?

Currently the SpanLink.trace_flags is set using TraceDigest.trace_flags. The trace_flag field in a trace_digest store sampling decisions propagated by tracecontext distributed headers. Therefore for non-remote spans and spans generated via datadog/b3 distributed tracing headers will have a trace_flag of nil even when a sampling decision exists.

SpanLink.trace_flag must have one of the following values:

• nil If there is NO sampling decision on the linked span
• 1 If the linked span is sampled
• 0 If the linked span is NOT sampled (span will be dropped by the agent)

Since trace_flag is ONLY a function of sampling priority, we can ignore TraceDigest.trace_flags and always use sampling_priority to compute the valid when a span link is generated.

[TonyCTHsu]

The code is generally good.

Question: When would digest be nil? I thinking about using null object pattern to avoid those nil check

[mabdinur]

The code is generally good.

Question: When would digest be nil? I thinking about using null object pattern to avoid those nil check

Digest shouldn't be nil. I'll make this change. We can make digest a required argument since the span_link class has not been released.