Consolidate thresholds for path build bias and use bias
In #8081 (moved), we added the ability to do separate accounting for path use bias (in addition to the existing path construction bias implemented previously). We likely want to keep these thresholds independent for a while, as the Tor network load adapts to ntor and onionskin changes, and we otherwise determine the relative failure rates users experience under normal circumstances for both construction and circuit use independently.
However, in the long run it will be easier to reason about the total impact on path selection that we allow through the combination of both types of failure as a single threshold. This will also simplify some of the logging and guard failure logic as well.
This might even be a tor 0.2.6.x thing, depending on how quickly the network transitions to ntor.