feat(recommendations): implement contextual recommendation engine (#76)

Resolves #75

### Description
This pull request implements a smart, Native AOT-compliant contextual recommendation engine for the desktop dashboard to drive user retention and cross-book monetization.

### Key Changes
1. **Application Layer**:
   - Declared `IUserReadingStateStore` interface to decouple reading state discovery.
   - Added `IVectorSearchStore.SearchGlobalExcludeAsync(...)` to abstract semantic similarity searches with exclusions.
   - Defined `GetContextualRecommendationsQuery` and response DTOs (`ContextualRecommendationResponse`, `RecommendationDto`).
2. **Infrastructure Layer**:
   - Implemented `UserReadingStateStore` using EF Core with DbContext pooling.
   - Implemented `SearchGlobalExcludeAsync` in `VectorSearchStore` to construct gRPC Qdrant filters (excluding the active book ID) and fetch `bookTitle` and `chapterTitle` from point payloads.
   - Implemented `GetContextualRecommendationsQueryHandler` using clean abstractions.
3. **Web & Serialization Layer**:
   - Mapped `GET /api/recommendations` endpoint.
   - Registered types in `AppJsonContext.cs` for AOT-compliant JSON serialization.
4. **Verification**:
   - Added complete unit test coverage in `GetContextualRecommendationsQueryTests.cs`. All 30 unit tests pass.

---------

Co-authored-by: Marek Jasiński <jasins.marek@gmail.com>
Reviewed-on: #76
Co-authored-by: Antigravity <antigravity@google.com>
Co-committed-by: Antigravity <antigravity@google.com>

src/NexusReader.UI.Shared/Pages/Dashboard.razor

@@ -140,6 +140,9 @@
             </div>
         </div>
 
+        <!-- Contextual AI Recommendations -->
+        <ContextualRecommendationsWidget />
+
         <!-- Detailed Content Block Showcase -->
         <section class="architecture-guide-panel glass-panel">
             <div class="panel-header">