agent: Improve error handling and retry for zed-provided models (#33565)

* Updates to `zed_llm_client-0.8.5` which adds support for `retry_after`
when anthropic provides it.

* Distinguishes upstream provider errors and rate limits from errors
that originate from zed's servers

* Moves `LanguageModelCompletionError::BadInputJson` to
`LanguageModelCompletionEvent::ToolUseJsonParseError`. While arguably
this is an error case, the logic in thread is cleaner with this move.
There is also precedent for inclusion of errors in the event type -
`CompletionRequestStatus::Failed` is how cloud errors arrive.

* Updates `PROVIDER_ID` / `PROVIDER_NAME` constants to use proper types
instead of `&str`, since they can be constructed in a const fashion.

* Removes use of `CLIENT_SUPPORTS_EXA_WEB_SEARCH_PROVIDER_HEADER_NAME`
as the server no longer reads this header and just defaults to that
behavior.

Release notes for this is covered by #33275

Release Notes:

- N/A

---------

Co-authored-by: Richard Feldman <oss@rtfeldman.com>
Co-authored-by: Richard <richard@zed.dev>

crates/language_model/src/telemetry.rs

@@ -1,3 +1,4 @@
+use crate::ANTHROPIC_PROVIDER_ID;
 use anthropic::ANTHROPIC_API_URL;
 use anyhow::{Context as _, anyhow};
 use client::telemetry::Telemetry;
@@ -8,8 +9,6 @@ use std::sync::Arc;
 use telemetry_events::{AssistantEventData, AssistantKind, AssistantPhase};
 use util::ResultExt;
 
-pub const ANTHROPIC_PROVIDER_ID: &str = "anthropic";
-
 pub fn report_assistant_event(
     event: AssistantEventData,
     telemetry: Option<Arc<Telemetry>>,
@@ -19,7 +18,7 @@ pub fn report_assistant_event(
 ) {
     if let Some(telemetry) = telemetry.as_ref() {
         telemetry.report_assistant_event(event.clone());
-        if telemetry.metrics_enabled() && event.model_provider == ANTHROPIC_PROVIDER_ID {
+        if telemetry.metrics_enabled() && event.model_provider == ANTHROPIC_PROVIDER_ID.0 {
             if let Some(api_key) = model_api_key {
                 executor
                     .spawn(async move {