Standardize on u64 for token counts (#32869)

Previously we were using a mix of `u32` and `usize`, e.g. `max_tokens:
usize, max_output_tokens: Option<u32>` in the same `struct`.

Although [tiktoken](https://github.com/openai/tiktoken) uses `usize`,
token counts should be consistent across targets (e.g. the same model
doesn't suddenly get a smaller context window if you're compiling for
wasm32), and these token counts could end up getting serialized using a
binary protocol, so `usize` is not the right choice for token counts.

I chose to standardize on `u64` over `u32` because we don't store many
of them (so the extra size should be insignificant) and future models
may exceed `u32::MAX` tokens.

Release Notes:

- N/A

crates/agent/src/thread.rs

@@ -272,8 +272,8 @@ impl DetailedSummaryState {
 
 #[derive(Default, Debug)]
 pub struct TotalTokenUsage {
-    pub total: usize,
-    pub max: usize,
+    pub total: u64,
+    pub max: u64,
 }
 
 impl TotalTokenUsage {
@@ -299,7 +299,7 @@ impl TotalTokenUsage {
         }
     }
 
-    pub fn add(&self, tokens: usize) -> TotalTokenUsage {
+    pub fn add(&self, tokens: u64) -> TotalTokenUsage {
         TotalTokenUsage {
             total: self.total + tokens,
             max: self.max,
@@ -396,7 +396,7 @@ pub struct ExceededWindowError {
     /// Model used when last message exceeded context window
     model_id: LanguageModelId,
     /// Token count including last message
-    token_count: usize,
+    token_count: u64,
 }
 
 impl Thread {
@@ -2769,7 +2769,7 @@ impl Thread {
             .unwrap_or_default();
 
         TotalTokenUsage {
-            total: token_usage.total_tokens() as usize,
+            total: token_usage.total_tokens(),
             max,
         }
     }
@@ -2791,7 +2791,7 @@ impl Thread {
         let total = self
             .token_usage_at_last_message()
             .unwrap_or_default()
-            .total_tokens() as usize;
+            .total_tokens();
 
         Some(TotalTokenUsage { total, max })
     }