Show nonclustered index count on modification operators in plan viewer

[erikdarlingdata]

Feature Request

When a modification operator (INSERT, UPDATE, DELETE) touches multiple indexes, display the count of nonclustered indexes maintained by that operator as a badge on the plan node.

Why

A single INSERT/UPDATE/DELETE on a table with many nonclustered indexes has hidden I/O cost — every row modification must maintain each nonclustered index. The clustered index (or heap) is implicit since it is the table, but the nonclustered maintenance overhead is not obvious from the operator name alone.

How it works in the plan XML

Modification operators contain multiple <Object> elements inside <Update> or <SimpleUpdate>, one per index touched:

<Update DMLRequestSort="false">
    <Object ... Index="[PK_Users_Id]" IndexKind="Clustered" Storage="RowStore" />
    <Object ... Index="[a0]" IndexKind="NonClustered" Storage="RowStore" />
    <Object ... Index="[a1]" IndexKind="NonClustered" Storage="RowStore" />
    <Object ... Index="[a2]" IndexKind="NonClustered" Storage="RowStore" />
</Update>

What to display

Count only IndexKind="NonClustered" objects. The primary object (clustered index or heap) is not counted since it's inherent to the operation.

This applies to both clustered tables and heaps — a Table Insert on a heap with 5 nonclustered indexes should show the same badge as a Clustered Index Insert with 5 nonclustered indexes.

Implementation

1. Parser: Count <Object> elements with IndexKind="NonClustered" inside <Update> / <SimpleUpdate> elements
2. Plan node model: Add NonClusteredIndexCount field
3. Plan viewer UI: Display as a badge on the operator node when count > 0

Parity

This feature is also tracked for PerformanceStudio: erikdarlingdata/PerformanceStudio#167