fix: more reliable span.sync determination, drop unused transaction.sync

[trentm]

Current span.sync is tracked via:

• the "before" async hook setting it false on the "active" span, and
• Instrumentation#bindFunction's wrapper setting it false, on the
  fair assumption that all usages of bindFunction are for async
  callbacks.
  The former has issues when there are multiple active spans within a
  single async task -- as is the case with Elasticsearch instrumentation
  (issue Async spans are being reported as sync #1996) and the aws-sdk instrumentations (which have manual
  workarounds).

This changes to set sync=false if the executionAsyncId() at end-time
is different than at start-time. This works for whatever asyncHooks
config var value.

One change in behaviour is that the value of span.sync is only
guaranteed to be accurate after span.end() is called. Given that
span.sync isn't a public field, I believe this is fine.

This change also drops the 'transaction.sync' fields. It was never
used by APM server.

Fixes: #1996
Fixes: #2292

Checklist

• Implement code
• Add tests
• Add CHANGELOG.asciidoc entry (waiting on 3.21.0 release first, 3.21.0 #2329)

[apmmachine]

💚 Build Succeeded

the below badges are clickable and redirect to their specific view in the CI or DOCS

Expand to view the summary

Build stats

• Start Time: 2021-09-20T17:15:10.889+0000

• Duration: 18 min 26 sec

• Commit: d512ebb

Test stats 🧪

Test	Results
Failed	0
Passed	20
Skipped	0
Total	20

Trends 🧪

[astorm]

Neat! (and smart!) -- overall I like this change and this seems like a better/more-accurate way of tracking whether a span has cross an asynchronous boundary (I'm increasingly skeptical of how useful this is to track for the general Node.js user -- but that's a separate issue :))

Happy to get on the approval train once the perf. and patch-async questions are worked through.

[astorm]

👍