[pg + pg-pool] Update DB semantic conventions

[dyladan]

open-telemetry/opentelemetry-specification#575

[naseemkullah]

This may fall out of scope but would it make sense to have a shared opentelemetry-semantic-conventions package that exports the enum of all attribute names and such? If so should this be package be hosted in the core repo?

[dyladan]

There is already such a package. https://github.com/open-telemetry/opentelemetry-js/tree/master/packages/opentelemetry-semantic-conventions

[naseemkullah]

perfect!

[naseemkullah]

@dyladan re: https://github.com/open-telemetry/opentelemetry-specification/blob/master/specification/trace/semantic_conventions/database.md#mysql

Please take note of span name and db statement.

Shall we proceed like as such? Currently span name is e.g.: mysql:query:SELECT, the value of db statement is that of span name and we have a custom attribute MYSQL_VALUES with the values. To adhere to spec, MYSQL_VALUES would be gone, they would be templated into the sql for db statement, and the non templated sql statement would be the span name.

If we proceed as per spec, should we worry about sanitization?

[dyladan]

We should follow the spec. The spec does not require sanitization, so I think we leave that as a backend concern. There is also an option to sanitize at the collector level, so I think we're ok.

[naseemkullah]

FYI I don't plan on tackling pg or pg-pool so those are up for grabs.