postgres_backend: don't print error stack traces

[erikgrinaker]

Problem

The use of :? when printing errors will dump stack traces, which can be overly dramatic and noisy for benign errors. For example:

2025-01-08T09:00:12.184800Z ERROR {cid=1168202 ttid=foo/bar application_name=Some("walreceiver")}: query handler for 'IDENTIFY_SYSTEM' failed: Timeline foo/bar was not found in global map
Stack backtrace:
   0: anyhow::error::<impl core::convert::From<E> for anyhow::Error>::from
             at /home/nonroot/.cargo/registry/src/index.crates.io-6f17d22bba15001f/anyhow-1.0.94/src/backtrace.rs:27:14
   1: <T as core::convert::Into<U>>::into
             at /rustc/90b35a6239c3d8bdabc530a6a0816f7ff89a0aaf/library/core/src/convert/mod.rs:759:9
   2: safekeeper::handler::SafekeeperPostgresHandler::handle_identify_system::{{closure}}::{{closure}}
             at /home/nonroot/safekeeper/src/handler.rs:403:46
   3: core::result::Result<T,E>::map_err
             at /rustc/90b35a6239c3d8bdabc530a6a0816f7ff89a0aaf/library/core/src/result.rs:856:27
   4: safekeeper::handler::SafekeeperPostgresHandler::handle_identify_system::{{closure}}
             at /home/nonroot/safekeeper/src/handler.rs:400:19
   5: <safekeeper::handler::SafekeeperPostgresHandler as postgres_backend::Handler<IO>>::process_query::{{closure}}
             at /home/nonroot/safekeeper/src/handler.rs:306:95
   6: <core::pin::Pin<P> as core::future::future::Future>::poll
             at /rustc/90b35a6239c3d8bdabc530a6a0816f7ff89a0aaf/library/core/src/future/future.rs:123:9
[50 more lines of stack]

Summary of changes

Don't print stack traces for errors.

[github-actions]

7264 tests run: 6905 passed, 0 failed, 359 skipped (full report)

---

Flaky tests (3)

Postgres 17

• test_pageserver_gc_compaction_smoke: release-x86-64

Postgres 15

• test_physical_replication_config_mismatch_max_locks_per_transaction: release-arm64

Postgres 14

• test_parallel_copy: release-arm64

Code coverage* (full report)

• functions: 31.2% (8411 of 26962 functions)
• lines: 48.0% (66779 of 139239 lines)

* collected from Rust tests only

---

_{The comment gets automatically updated with the latest test results
2d36a1d at 2025-01-08T13:20:29.984Z :recycle:}

[arpad-m]

I have found backtraces to be useful, not in the walreceiver, but in other areas, to quickly identify where an error is coming from. Given that this is a place that logs at the ERROR log level, we should reach it rarely.

If the error is benign, we should catch it earlier and downgrade it to warning or info level. We've done so in the past for many other errors.

[erikgrinaker]

The Safekeeper logs are full of these kinds of errors, happens all the time due to various races. I think the error type is opaque here (it's just an anyhow::Error). But sure, I'll see if I can tweak it.