fix: don't silence errors on span update calls

[zachdaniel]

This addresses some but not all of the issues in #65. I'm still not sure how we should return from start_trace start_span finish_trace finish_span and the continue_* functions if given invalid updates. Its been discussed in slack, and we probably shouldn't just skip starting the span, and we also should raise an error. Likewise, returning {:error, error} but still starting the trace/span might be wrong. Really not sure of the best path forward there.

[zachdaniel]

I've settled on accepting that some calls silently ignore errors, and have illustrated that behavior in tests and in documentation.

[coveralls]

Pull Request Test Coverage Report for Build 4ddad899-ed81-48d2-85a9-a30ba6d98f6c

• 13 of 13 (100.0%) changed or added relevant lines in 1 file are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage increased (+0.9%) to 83.399%

---

Totals
Change from base Build 971d501c-6d07-4be5-bc75-90c95cf7afb9:	0.9%
Covered Lines:	211
Relevant Lines:	253

---

💛 - Coveralls

[GregMefford]

LGTM in general, just a bunch of nitpicks. 😁

[GregMefford]