Fix compute nodes stuck at bidding

[wdbaruni]

A compute node can bid on a job even after another bid was accepted and results were published, which will result in that compute node getting stuck in bidding state with capacity reserved as the requester node will ignore the bid and won't reject it. There are couple bugs that were fixed to mitigate this issue:

1. Fix a bug in the compute node that checks if a shard has reached capacity before bidding on. Previously it checks if there are enough nodes in bid accepted state, but it should also check if any state post bid accepted, such as completed!
2. Requester node now returns a InvalidRequest event back to the compute node if it bids on a job that is no longer biddable (e.g. completed or reached capacity), or if the job doesn't exist (e.g. requester node restarted)

A long term solution is to implement timeouts in each state to let the compute node fail rather than getting stuck, if for example the requester node is no longer available. This is tracked in #663