Allow subclassing SpanContext by other packages

[tylerbenson]

Having abstract methods with default visibility makes the class not able to be subclassed by other packages.
Changing to protected to maintain the intended restrictions from #1594, but allowing it to be subclassed.

[codecov]

Codecov Report

Merging #1940 into master will increase coverage by 0.04%.
The diff coverage is 100.00%.

@@             Coverage Diff              @@
##             master    #1940      +/-   ##
============================================
+ Coverage     86.63%   86.67%   +0.04%     
- Complexity     1501     1502       +1     
============================================
  Files           184      184              
  Lines          5662     5664       +2     
  Branches        582      582              
============================================
+ Hits           4905     4909       +4     
+ Misses          551      550       -1     
+ Partials        206      205       -1     

Impacted Files	Coverage Δ	Complexity Δ
.../api/baggage/propagation/W3CBaggagePropagator.java	93.18% <ø> (ø)	15.00 <0.00> (ø)
...ava/io/opentelemetry/api/trace/PropagatedSpan.java	79.16% <ø> (ø)	17.00 <0.00> (ø)
...n/java/io/opentelemetry/api/trace/SpanContext.java	89.47% <ø> (ø)	14.00 <0.00> (ø)
.../java/io/opentelemetry/context/DefaultContext.java	100.00% <ø> (ø)	7.00 <0.00> (ø)
...ain/java/io/opentelemetry/context/LazyStorage.java	88.46% <ø> (ø)	9.00 <0.00> (ø)
...va/io/opentelemetry/opentracingshim/TraceShim.java	100.00% <100.00%> (ø)	4.00 <1.00> (+1.00)
...y/sdk/metrics/aggregator/DoubleMinMaxSumCount.java	100.00% <0.00%> (+4.08%)	6.00% <0.00%> (ø%)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update c6f111a...89d1ca3. Read the comment docs.

[jkwatson]

Why do you need to create a non-AutoValue subclass of SpanContext? I think this is probably not a good idea, since we do not want the SpanContext to end up mutable. If we allow subclassing, we cannot guarantee that.

Is there another way to solve your use-case that doesn't open up the possibility of breaking things?

[Oberon00]

@tylerbenson The spec says that the SpanContext functionality SHOULD NOT be overridable, so I don't think we should do that.

[tylerbenson]

Closing in favor of #1935.