atomic gas charging

[gballet]

This is an experiment to look into charging the gas "atomically" instead of on a per-leaf basis.

Differences are:

• if there isn't enough gas to pay for the proving of all the leaves in a "cluster", then none of the leaves are added to the witness.
• it's simpler, albeit more verbose: all corner cases where there is enough gas for one leaf but not another one, is
• the witness is smaller
• it will make the geth team happy, and it might even get merged some day!
• it will require a spec change

The principle is simple:

• Decouple the gas computation from the "touching"
• Only return the "wanted" part from the gas computation
• If an overflow is detected, just return ErrOutOfGas, which will cause all the remaining gas to be consumed - this simplifies all the issues we had of knowing when there isn't enough gas
• touching does it blindly

TODO

• run execution spec tests and confirm that it only fails on the tests that correspond to partial gas charging
• update the spec
• check the len(ret) > 0 part
• get TestProcessVerkle to pass
• shadowfork holesky it seems that the fork itself has been broken due to lack of maintenance, if this is broken by it (it shouldn't) then it will be fixed during the rebase
• benchmark how much is lost when passing variadic arguments, and if this is optimized out by using the stack this is no longer needed as the API was changed to use a range instead.

[jsign]

availableGas parameter is redundant