Hedging: Fixes rare NullRefreneceException in Headers Clone

[NaluTripician]

Pull Request Template

Description

There is a rare case where a NullReferenceException can occur when cloning headers in the Headers class, in request hedging scenarions

System.ArgumentNullException: Value cannot be null. (Parameter 'value')
at Microsoft.Azure.Cosmos.Headers.Clone()
at Microsoft.Azure.Cosmos.RequestMessage.Clone(ITrace newTrace, CloneableStream cloneContent)
at await Microsoft.Azure.Cosmos.CrossRegionHedgingAvailabilityStrategy.CloneAndSendAsync(?)
at await Microsoft.Azure.Cosmos.CrossRegionHedgingAvailabilityStrategy.ExecuteAvailabilityStrategyAsync(?)
at await Microsoft.Azure.Cosmos.Handlers.RequestInvokerHandler.SendAsync(?)
at await Microsoft.Azure.Cosmos.Handlers.RequestInvokerHandler.SendAsync(?)
at await Microsoft.Azure.Cosmos.ContainerCore.ProcessItemStreamAsync(?)
at await Microsoft.Azure.Cosmos.ContainerCore.ReadItemStreamAsync(?)
at await Microsoft.Azure.Cosmos.ClientContextCore.RunWithDiagnosticsHelperAsync(?)
at await Microsoft.Azure.Cosmos.ClientContextCore.OperationHelperWithRootTraceAsync(?)

This pull request introduces a null check in the Clone method of the Headers class and adds new unit tests to verify the cloning functionality, including handling of null values.

Enhancements to Headers class cloning:

• Microsoft.Azure.Cosmos/src/Headers/Headers.cs: Added a null check in the Clone method to ensure that headers with null values are not added to the cloned headers.

New unit tests for Headers class:

• Microsoft.Azure.Cosmos/tests/Microsoft.Azure.Cosmos.Tests/HeadersTests.cs: Added TestHeadersClone to verify that all keys and values are correctly cloned.
• Microsoft.Azure.Cosmos/tests/Microsoft.Azure.Cosmos.Tests/HeadersTests.cs: Added TestNullHeadersClone to ensure that headers with null values are handled correctly during cloning.

Type of change

Please delete options that are not relevant.

• [] Bug fix (non-breaking change which fixes an issue)

Closing issues

To automatically close an issue: closes #5093

[Pilchie]

Is it expected that there be nulls in there? Instead of (or as well as) this should we be investigating how the null got in there?

[kirankumarkolli]

I assumed that it's a concurrency issue, otherwise +1

[Pilchie]

What concurrency issue? Someone is writing to the headers collection on one thread while we're reading? That's not safe.

[NaluTripician]

Fixed in: #5189