feat!: gateway: fix rate limiting, better stateful handling

[rvagg]

Ref: filecoin-project/go-jsonrpc#118
Ref: #11589
Ref: #11153

---

API changes and breakage:

• Plain HTTP connections now can't use stateful ethereum APIs (they couldn't really before, but they allowed addition of filters and subs in a dangerous way), you have to be using WebSockets
• gateway.NewNode now only takes 1 argument (the api) and a list of options to override defaults
• gateway.NewRateLimiterHandler and gateway.NewConnectionRateLimiterHandler have been replaced with gateway.NewRateLimitHandler.
• The handlers returned by both gateway.NewRateLimitHandler and the primary gateway.Handler return an http.Handler augmented with a Shutdown(ctx) method to be used for graceful cleanup of resources.
• The default for the config option Events.FilterTTL has been reduced from 24h to 1h. It's a measure of "how long since this thing was last used", and 24h is way too long for that IMO.

Fix:

• --per-conn-rate-limit was previously applied as a global rate limiter,
  effectively making it have the same impact as --rate-limit. This change fixes
  the behaviour such that --per-conn-rate-limit is applied as a API call
  limiter within a single connection (i.e. a WebSocket connection). The rate
  is specified as tokens-per-second, where tokens are relative to the expense
  of the API call being made.

Improve:

• Auto-clean-up of stateful resources when a WebSockets connection ends, so we close down subscriptions and remove filters if the user didn't do so themselves.
• Added --eth-max-filters-per-conn to override the default (16) and it now applies to the combined total of eth subs and filters, not separately
• Replace the --conn-per-minute limit mechanism to a standard rate.Limiter instead of the custom rate limit implementation. There is a slight change of behaviour with this in that I allow a burst, which the original doesn't. It also has a different cleanup mechanism with a separate goroutine dedicated to this task instead of a goroutine per request that decrements and may cleanup if zero.

[masih]

Two potential blockers:

• DoS attack vector due to unbounded limiters per host.
• Repeated timer instantiation in loop instead of reusing ticker.

Other than that, I left a bunch of comments/suggestions.

Thank you for working on this @rvagg 🍻

[rvagg]

@masih thanks for the thorough review. I've gone through all your feedback and marked as resolved the easy bits here but left as unresolved the things that may either need follow-up or have responses to questions. Would you mind having another look please?

[rvagg]

I've messed something up in rate limiting in my latest round of changes, TestGatewayRateLimits is failing now. I need to fix that first.

[rvagg]

@masih || @aarshkshah1992 would you mind having one last look over this before I merge please? @masih the final commit should be the delta between your review and now and contains mostly things you raised.