Add extended version of puffin::profile_scope!() (#213)

# Checklist

* [X] I have read the [Contributor Guide](../CONTRIBUTING.md)
* [X] I have read and agree to the [Code of
Conduct](../CODE_OF_CONDUCT.md)
* [X] I have added a description of my changes and why I'd like them
included in the section below

# Description of Changes

Add extended version of `puffin::profile_scope!()` that expands to an
expression instead of variable

- This allows for profiling scopes to persist for custom scopes, instead
of being limited to block scopes
- Includes doctests and real-world example


# Related Issues

As per request from @chrisduerr in #194

puffin/src/lib.rs

@@ -231,24 +231,47 @@ macro_rules! profile_function {
     };
 }
 
-#[allow(clippy::doc_markdown)] // clippy wants to put "MacBook" in ticks 🙄
+
 /// Profile the current scope with the given name (unique in the parent scope).
+/// 
+/// This macro is identical to [profile_scope], except that it expands to the expression
+/// containing the profiling scope, as opposed to [profile_scope] which expands to a
+/// variable (which cannot be accessed due to macro hygiene).
 ///
-/// Names should be descriptive, ASCII and without spaces.
-///
-/// Example: `profile_scope!("load_mesh");`.
+/// This allows for profiling scopes to persist for a custom duration.
+/// 
+/// # Example
+/// 
+/// ```rust
+/// # use std::iter::FromIterator as _;
+/// # 
+/// # pub mod rayon { pub mod prelude {
+/// #     pub fn for_each_init<T, I>(vec: &std::vec::Vec<T>, init: fn() -> I, body: fn ((I, T)) -> ()) {
+/// #     }
+/// # } }
+/// #
+/// let some_large_vec = Vec::from_iter(0..1000);
 ///
-/// An optional second argument can be a string (e.g. a mesh name) to help diagnose what was slow.
-/// Example: `profile_scope!("load_mesh", mesh_name);`
-///
-/// Overhead: around 54 ns on Macbook Pro with Apple M1 Max.
+/// // Use rayon's parallel for loop over our large iterator
+/// rayon::prelude::for_each_init(
+///         &some_large_vec,
+///         // This gets called to init each work segment, and is passed into the calls
+///         // Rayon keeps the profiling scope stored for the entire duration of the work segment
+///         // So we can track the entire segment as one, instead of each loop iteration
+///         || puffin::profile_scope_custom!("rayon_work_segment"),
+///         |((_profiler_scope), i)| {
+///             // All calls here gets profiled on the same scope
+///             println!("{i}")
+///         },
+/// );
+/// ```
 #[macro_export]
-macro_rules! profile_scope {
+macro_rules! profile_scope_custom {
     ($name:expr) => {
-        $crate::profile_scope!($name, "");
+        $crate::profile_scope_custom!($name, "")
     };
-    ($name:expr, $data:expr) => {
-        let _profiler_scope = if $crate::are_scopes_on() {
+    ($name:expr, $data:expr) => {{
+        if $crate::are_scopes_on() {
             static SCOPE_ID: std::sync::OnceLock<$crate::ScopeId> = std::sync::OnceLock::new();
             let scope_id = SCOPE_ID.get_or_init(|| {
                 $crate::ThreadProfiler::call(|tp| {
@@ -264,7 +287,28 @@ macro_rules! profile_scope {
             Some($crate::ProfilerScope::new(*scope_id, $data))
         } else {
             None
-        };
+        }
+    }};
+}
+
+#[allow(clippy::doc_markdown)] // clippy wants to put "MacBook" in ticks 🙄
+/// Profile the current scope with the given name (unique in the parent scope).
+///
+/// Names should be descriptive, ASCII and without spaces.
+///
+/// Example: `profile_scope!("load_mesh");`.
+///
+/// An optional second argument can be a string (e.g. a mesh name) to help diagnose what was slow.
+/// Example: `profile_scope!("load_mesh", mesh_name);`
+///
+/// Overhead: around 54 ns on Macbook Pro with Apple M1 Max.
+#[macro_export]
+macro_rules! profile_scope {
+    ($name:expr) => {
+        $crate::profile_scope!($name, "");
+    };
+    ($name:expr, $data:expr) => {
+        let _profiler_scope = $crate::profile_scope_custom!($name, $data);
     };
 }