search: ClassCastException when a facet returns zero buckets

## Summary

Any `search` call with `facetFields` that produces **zero facet buckets** fails with:

```
class java.util.ArrayList cannot be cast to class org.apache.solr.common.util.NamedList
```

This is not specific to a field type or a collection — the same field succeeds or fails purely
depending on whether the query matched anything.

## Reproduction

Solr 9 in SolrCloud mode, a collection `shows` with a single-valued `string` field `platform`
(docValues on), 61 documents indexed.

**A — facet has buckets → works**

```json
{"collection": "shows", "query": "*:*", "facetFields": ["platform"], "rows": 0}
```
```
{"numFound":61,...,"facets":{"platform":{"Netflix":20,"HBO Max":7,...}}}
```

**B — same field, filter matches nothing → throws**

```json
{"collection": "shows", "query": "platform:NoSuchPlatform", "facetFields": ["platform"], "rows": 0}
```
```
class java.util.ArrayList cannot be cast to class org.apache.solr.common.util.NamedList
```

Verified end-to-end over MCP STDIO against a published container image, and independently over
the HTTP transport.

## Root cause

Solr serializes an empty facet field as an empty JSON **array**:

```json
"facet_counts": { "facet_fields": { "platform": [] } }
```

`JsonResponseParser.isFlatNamedList` rejects zero-length arrays
([`JsonResponseParser.java:157`](https://github.com/apache/solr-mcp/blob/main/src/main/java/org/apache/solr/mcp/server/config/JsonResponseParser.java#L157)):

```java
int size = arrayNode.size();
if (size == 0 || size % 2 != 0)
return false;
```

so `convertArray` falls through to the plain-list branch and returns an `ArrayList`. SolrJ's
`QueryResponse.getFacetFields()` then casts that value to `NamedList` and throws.

The heuristic cannot be fixed by simply allowing `size == 0`: a bare `[]` is genuinely ambiguous
between an empty facet NamedList and an empty plain list, and the parser has no context to tell
them apart.

## Suggested fix

Request `json.nl=map` explicitly rather than relying on Solr's `flat` default. Facet fields then
arrive as JSON objects:

```json
"facet_fields": { "platform": {"Netflix": 20} } // non-empty
"facet_fields": { "platform": {} } // empty
```

Both map unambiguously onto `NamedList` through the existing `convertObject` path, which removes
the flat-array heuristic and this whole class of ambiguity rather than patching the symptom.

## Impact

A user asking a perfectly ordinary question — *"break these results down by platform"* where the
filter happens to match nothing — gets a raw Java exception instead of an empty result set. Zero
matches is a normal outcome of a search, not an error condition.

## Environment

- `main` @ `a84033b`; `JsonResponseParser.java` unmodified at that commit
- Solr 9 (`solr:9-slim`), SolrCloud mode
- solr-mcp 1.0.0-SNAPSHOT, Spring Boot 3.5.14

Start with JsonResponseParser.java, especially isFlatNamedList and the conversion path described in the report, then trace how search requests set Solr response parameters. Reproduce the supplied zero-match facet query over Solr 9 and verify that requesting json.nl=map makes empty facet_fields parse as NamedList and return an empty result instead of throwing.

Written by the indexing model from the issue text.