Filtering (for traces) should allow to skip unaffected spans

[pilhuhn]

A filter expression like

filter:
  application:
    server.port:
      match: 8787

will only generate a span if the server port is 8787. This has the effect, that internal spans for e.g. for DB-calls triggered from a process are not reported, as the server port for the (targeted) DB is different.
Of course I could give multiple ports to match, but that requires me to know the internals of observed software.

The filter-matching should probably be limited to e.g. only http-servers and ignore SQLClient spans in the matching.

[grcevski]

@mariomac you are expert in this area, are we able to apply the filtering to a specific series, like http_server_request_seconds...?