KeepAlive for @sentry/node transports

[rhcarvalho]

In #1795 the transports changed from keepAlive: true to keepAlive: false. The reason had to do with memory leaks in old versions of Node.

keepAlive: false is undesirable because it incurs more overhead (re-)establishing TCP connections instead of keeping idle connections in the pool.

https://nodejs.org/dist/latest-v12.x/docs/api/http.html#http_class_http_agent

I propose we set keepAlive: true, at least for recent versions of Node.

[smeubank]

@JonasBa should we keep this issue open? is the desire to change the default setting or really to apply some improvements with regards to which setting is placed? Or dependent on which feature is being used? As with profiling we may have more power to apply what the standard should be

[JonasBa]

@smeubank, I would love to flip this, but I think the risk of introducing a memory leak for users on outdated node versions is still pretty large. The leak was fixed somewhat recently here, so any outdated v12 or v14 would still be impacted.

Looking at https://endoflife.date/nodejs, v14 ends life support in 4 months, so I think we should enable it by default then (or conditionally via process.versions.node if we want to err on the safe side).

That said, we did add the keepAlive option to the transporter so users can override it keepAlive cc @rhcarvalho.

[smeubank]

should we put this on the radar for a profiling GTM plan? Happy if we want to reopen the issue

[JonasBa]

@smeubank I would treat this as an isolated improvements, I dont think we'll have the bandwidth to run a benchmark and put some numbers on this, I'd rather just flip once v14 support ends.