Bring current protocol costs in-line with current parameter estimator implementation

[matklad]

Historically, we only measured number of CPU instructions executed when calculating costs. This is the implementation which was used to derive the current set of costs.

Then, in #3771, we implemented the logic to also account for the cost of IO operations (by measuring the total number of bytes read/written, and adding them with some coeficient to the instruction counter).

Crucially though, we didn't actually update the costs after that work (the PR to do this isn't merged: #3818). That means that, if you run parameter estimeator today, you get a wildly different costs from the status quo. Current costs doesn't account for IO, but the estimator does.

This is a significant amount of technical dept: if we want to change some cost today, we can't just run parameter estimator and look at the diff, as that would be comparing apples and oranges.

[stale]

This issue has been automatically marked as stale because it has not had recent activity in the last 2 months.
It will be closed in 7 days if no further activity occurs.
Thank you for your contributions.

[stale]

This issue has been automatically marked as stale because it has not had recent activity in the last 2 months.
It will be closed in 7 days if no further activity occurs.
Thank you for your contributions.