Improve handling of unsupported subgraph indexing networks

[fordN]

The UX for unsupported subgraphs is leaving a lot to be desired right now.

Why?

• We do not support indexing rules that specify the indexed network, so often subgraphs are identified for allocation that the graph-node cannot support.
• Deployments for unsupported subgraphs are attempted repeatedly and included in prometheus error metrics leading to false alarms in alerting and monitoring

These UX issues are amplified on a larger scale indexer like some of the E&N indexers. If subgraph deployments are hand selected by the indexer these pain points will likely not be exposed.

[fordN]

Brainstorming some solutions:

• Indexing rules could specify an indexing network which would allow the indexer to be selective about which indexing networks they support.
• The indexer-agent could query the graph-node instances to see what chains are supported and automatically disqualify subgraph deployments that sync an unsupported chain or feature.
• Error handling could be improved to simply make the unsupported network errors less chatty.

[fordN]

Related to graphprotocol/allocation-optimizer#42

[alex-pakalniskis]

#836

[alex-pakalniskis]

#894