fix: remove double-logging from capture_exception -- saves log ingestion cost

capture_exception() was calling logger.error() internally, but every
caller (50 call sites) also calls logger.error() before/after it.
Result: every error printed to stdout twice, doubling log volume for
errors in production log aggregators.

Now capture_exception() only sends to Sentry. Callers handle their
own logging. Fallback logger.error() only fires when Sentry SDK is
not installed (so errors aren't silently lost in local dev).

284 tests pass.

Author: DevanshuNEU

backend/services/observability.py

@@ -155,26 +155,20 @@ def add_breadcrumb(message: str, category: str = "custom", level: str = "info",
 
 def capture_exception(error: Exception, **context):
     """
-    Capture exception with additional context.
+    Capture exception to Sentry with additional context.
     
-    Args:
-        error: The exception to capture
-        **context: Additional context to attach
+    Does NOT log to stdout -- callers are responsible for logging.
+    This avoids double-logging when callers do logger.error() + capture_exception().
     """
     try:
         import sentry_sdk
         with sentry_sdk.push_scope() as scope:
             for key, value in context.items():
                 scope.set_extra(key, value)
             sentry_sdk.capture_exception(error)
-        
-        # Also log it
-        logger.error(
-            f"Exception captured: {type(error).__name__}: {str(error)}",
-            **context
-        )
     except ImportError:
-        logger.error(f"Exception: {error}", **context)
+        # No Sentry -- log as fallback so errors aren't silently lost
+        logger.error(f"Exception (no Sentry): {type(error).__name__}: {error}", **context)
 
 
 def capture_message(message: str, level: str = "info", **context):