Initialize OOM protection in GrpcQueryServer

[jitendrakr88]

Description

When OOM protection is enabled, the grpc streming queries always return 0 rows. This issue has been observered with spark connector as well as trino connector (reported by another user).

from pyspark.sql.functions import col

cutoff_seconds_since_epoch = 1741169418

df = spark \
    .read \
    .format("pinot") \
    .option("controller", "<host>:<port>") \
    .option("table", "<table>") \
    .option("useGrpcServer", "true") \
    .option("tableType", "REALTIME") \
    .option("segmentsPerSplit", 1) \
    .load() \
    .filter(col("secondsSinceEpoch") >= cutoff_seconds_since_epoch)

print("Number of rows: " + str(df.count())) 

Observation:

• Number of rows returned is always zero if OOM is enabled on server tenant
• After disabling OOM, the grpc streaming query works fine.
• After disabling useGrpcServer option, the query works fine with http transport.

Solution

This PR fixes the issue by initialising the missing tracing context in GrpcQueryServer for OOM protection.

[cbalci]

This test is fine, but can we write a test case which actually validates the gRPC query server can execute queries?

The reason I'm insistent on the test is that the Grpc transport is secondary to the the HttpServer in terms of importance and when folks do sweeping changes (like the tracing setup), regressions to the gRPC server slip unnoticed.

[Jackie-Jiang]

(minor) Please fix the format