Improve performance of filtering merge requests by an environment

What does this MR do?

When filtering merge requests by an environment name, the index used by DeploymentMergeRequest.deployed_to() is not ideal. We can improve performance by adding the following AND condition:

AND environments.project_id = merge_requests.target_project_id

This condition forces the query to use the index on merge_requests (project_id, name), instead of using the slower index (name varchar_pattern_ops).

This improves performance of queries such as this:

SELECT merge_requests.*
FROM merge_requests
WHERE merge_requests.target_project_id = 278964
AND EXISTS (
    SELECT 1
    FROM deployment_merge_requests
    INNER JOIN deployments ON deployments.id = deployment_merge_requests.deployment_id
    INNER JOIN environments ON environments.id = deployments.environment_id
    WHERE deployment_merge_requests.merge_request_id = merge_requests.id
    AND environments.name = 'gstg'
    AND environments.project_id = merge_requests.target_project_id
)
ORDER BY merge_requests.id DESC
LIMIT 100;

For more information, refer to #270068 (comment 432945738).

Does this MR meet the acceptance criteria?

Conformity

• Changelog entry Feature is behind an unreleased feature flag
• Documentation (if required)
• Code review guidelines
• Merge request performance guidelines
• Style guides
• Database guides
• Separation of EE specific content