fix: break graph api notifiers across a sync map

[willscott]

Limits how many other exchanges / notifiers are notified.

Aims to fix #92

[lidel]

@willscott @aarshkshah1992 @aschmahmann @hacdias FYSA
it is now possible to test PRs on staging env without having to merge them to main

I've added support for staging branch – see https://github.com/ipfs/bifrost-gateway#docker, tldr:

When you want to test some code on staging (bifrost-stage1-ny), make sure code is in staging branch (ok to force-push) , docker-image build finished, and then manually ssh to bifrost-stage1-ny and use /data/update-bifrost.sh to deploy version you want. ( staging-latest or staging-YYYY-DD-MM-GITSHA if you want to pin a specific commit)

Demo with this PR:

1. force-pushed contents of this PR to staging
2. CI built staging-latest and published it at https://hub.docker.com/r/ipfs/bifrost-gateway/tags
3. Updated /data/update-bifrost.sh to use staging-latest and run it, you can see it runs the last commit from this PR:

2023/04/25 14:50:40 Starting bifrost-gateway 2023-04-25-8e1661e

[lidel]

Due to https://github.com/protocol/bifrost-infra/pull/2492 we had to restart the experiment:

2023/04/25 20:14:30 Starting bifrost-gateway 2023-04-25-8e1661e

[aarshkshah1992]

@lidel Which file should I change if I want to change some environment/config variables for Bifrost ?
When will they get overwritten by CI ?

[willscott]

@aarshkshah1992 https://github.com/protocol/bifrost-infra/blob/master/ansible/inventories/bifrost-gw/group_vars/gateway_bifrost_stage.yml#L43 is probably what you're after?

[lidel]

@aarshkshah1992 /data/update-bifrost.sh sets values ad-hoc, if you want to keep them around across automated restarts, you need to use file linked by @willscott

[lidel]

I think this PR did remove the perf. erosion over ~2h time window described in #92.

We had a pretty long (~12h) window running 8e1661e, and it maintained correctness over time:

Let's merge this, so we move focus to the next dragon.
@aschmahmann can revisit this as part of #75