Algod: Replace logic.DebuggerHook with logic.EvalTracer interface

[jdtzmn]

Summary

This PR upgrades the existing logic.DebuggerHook interface to be more powerful & lightweight. To be specific:

• Introduce new logic.EvalTracer interface and rename existing logic.DebuggerHook interface to logic.Debugger to be more intuitive.
• New EvalTracer hooks are introduced for before/after transaction groups, before/after transactions, and after TEAL opcodes.
• The TEAL evaluator will carry forward the debugger into any inner transaction evaluations.
• The EvalContext.refreshDebugState function which used to be called before every TEAL op has been removed. This function copied and base64 encoded the entire stack and scratch space every time it was invoked, which is potentially expensive and not needed for the eventual simulation trace.

A new logic.MakeEvalTracerDebuggerAdaptor adaptor function has been created so that existing logic.Debugger implementors (e.g. tealdbg and the dryrun endpoint) can be ran as logic.EvalTracer with the same behavior.

What's Next

By itself this PR doesn't activate any new behavior. Instead, it defines the tracer hooks that will be necessary for the following:

• Near term goal: return ApplyData information from simulated transactions
  • It's not necessary to use a tracer to access this info if the transactions succeed, however if a transaction fails, the block evaluator will not return the partially-completed ApplyData. To solve this without modifying the block evaluator, we can use a tracer to save a copy of the ApplyData after each operation. That's what the follow-up PR Algod: Additional simulation result information #4439 does.
• Longer term goal: enable simulation trace collection

Test Plan

• The old debugger tests still pass, now using logic.MakeEvalTracerDebuggerAdaptor
• Test BlockEvaluator.TransactionGroup calls tracer hooks correctly
• Test verify.TxnGroupWithDebugger calls tracer hooks correctly