ref: Ensure all aliases have a corresponding file

[lcian]

Everything that is present in alias/replacement should be an existing attribute.
This adds a test to check for that.

Also removes the following conventions that were detected by the test:

• "db.collection"
  Added in aa49460 by @AbhiPrasad

  • Action: removed.
  • Reason: I have checked getsentry/sentry, Relay, OTEL and this doesn't exist. There's just db.collection.name which is correctly represented as an attribute separate to this one.

• http.client_id, alias of client.address
  Added in 337143d by @AbhiPrasad

  • Action: removed.
  • Reason: this looks like it was a typo of http.client_ip.

• "http.response.header['content-length']",

  • Action: removed.
  • Reason: This was marked as an alias of http.response.header.content-length.
    I've removed it because it would introduce an additional special case when writing files to disk to handle the [], or we would have to change the [key] format.
    It's not present in sentry/Relay/OTEL.
    Note that, while the OTEL spec says any UTF-8 is a valid span attribute name, in practice OTEL tooling for semconv doesn't support this form, allowing only for . and _ as special characters, so if we eventually want to use that tooling that's another reason to avoid attribute names like this.

• "sentry.nextjs.function.route", "sentry.nextjs.function.type"
  These 2 have been added here feat: Add attribute for Next.js SSR functions #54

  • Action: removed.
  • Reason: if you follow the trail in the JS repo you can see that no release of the SDK used the wrong names (without ssr), so we can get rid of them.

[gggritso]

@lcian thanks for putting this together! Removing the buffer attributes in #110 👍🏻