Cache inflight current chain-head requests

[thewhaleking]

Adds new functionality to CachedFetcher that will not cache the results of a request if specified cache_results=False.

This may seem to defeat the point, but in fact it does not. This allows using the current request caching, without needing to cache the result.

For example, we wouldn't want to cache get_block_hash(None) because the block hash changes often. But if we can utilise the same inflight request for two concurrent requests, we can stop writing:

block_hash = await substrate.get_chain_head()
q1, q2 = await asyncio.gather(
    substrate.query(bar, foo, block_hash=block_hash),
    substrate.query(bat, baz, block_hash=block_hash)
)

and instead write the much simpler:

q1, q2 = await asyncio.gather(
    substrate.query(bar, foo),
    substrate.query(bat, baz)
)

as both requests internally fetch the chain head, and will utilise the same head fetch request with this new functionality. (See tests/unit_tests/asyncio_/test_substrate_interface_async_unit.py:test_concurrent_block_hash_none_dedups_chain_head_request)

[thewhaleking]

Fix for flaky bittensor e2e tests:
RaoFoundation/bittensor#3366
RaoFoundation/bittensor#3367

[basfroman]

btw, you might not have given it much thought, but this fixes the issue where a result of 0 was previously not cached. Now, it is.
Great!

[thewhaleking]

btw, you might not have given it much thought, but this fixes the issue where a result of 0 was previously not cached. Now, it is. Great!

Ha. I totally didn't even realise that.