feat: vm.pauseTracing + vm.resumeTracing

[klkvr]

Motivation

ref foundry-rs/forge-std#505 (comment), and I've also seen a request for this in one of the telegram chats recently (can't find now)

Adds pauseTracing and resumeTracing cheats which allow hiding parts of the call trace.

Solution

• Extend CheatcodeExecutor with fn tracing_inspector(&mut self) -> Option<&mut TracingInspector> allowing to manipulate traces and tracing configuration from cheatcodes. This should also be useful for feat(cheatcode): startDebugTraceRecording and stopDebugTraceRecording for ERC4337 testing #8571 and feat(cheatcodes): mark unmatched expectedEmits as unemitted #8686
• Record ranges of ignored traces into the following struct:

  foundry/crates/cheatcodes/src/utils.rs

  Lines 11 to 23 in 556eb84

  	/// Contains locations of traces ignored via cheatcodes.
  	///
  	/// The way we identify location in traces is by (node_idx, item_idx) tuple where node_idx is an
  	/// index of a call trace node, and item_idx is a value between 0 and `node.ordering.len()` where i
  	/// represents point after ith item, and 0 represents the beginning of the node trace.
  	#[derive(Debug, Default, Clone)]
  	pub struct IgnoredTraces {
  	/// Mapping from (start_node_idx, start_item_idx) to (end_node_idx, end_item_idx) representing
  	/// ranges of trace nodes to ignore.
  	pub ignored: HashMap<(usize, usize), (usize, usize)>,
  	/// Keeps track of (start_node_idx, start_item_idx) of the last `vm.pauseTracing` call.
  	pub last_pause_call: Option<(usize, usize)>,
  	}

  
  Another approach could be to remove trace nodes and items through a mutable reference, though we're using traces for debugger and gas reports, so keeping an instance of complete trace is useful even if it was partially ignored
• Changed type Traces = Vec<(TraceKind, CallTraceArena)> to Vec<(TraceKind, SparsedTraceArena)> where SparsedTraceArena keeps data about ignored traces and when being printed, constructs a copy of call trace arena without ignored items.
• Also included drive-by fix for CheatcodeExecutor::exec_create to support EOF bytecode

This already works pretty much reliably, though it messes with TracingInspector internals and thus there's at least one edge case I can think of related to --decode-internal:

function _someInternalFn() internal {
    vm.pauseTracing();
}

function test() public {
    _someInternalFn();
}

Such test will result in a panic during printing if run with --decode-internal as pauseTracing would result in end of decoded trace step being removed, causing issues in TraceWriter

We might want to introduce support for this feature on revm-inspectors level to address this and to also support more verbose output for ignored traces, e.g.:

├─ [0] VM::pauseTracing()
├─ [...30 collapsed traces...]
├─ [0] VM::unpauseTracing()

[mattsse]

I think this is reasonable, this feature isn't too invasive and still easy to follow

pedantic doc nits