[HUD] SGLang vs VLLM comparison dashboard (#7101)

This PR consists of the changes of adding a new dashboard which would be
used for comparing the results between SGLang and vLLM for all the
different metrics.

**Changes**:
1. Added a new property `repos` which is used for handling the
implementation when comparing two or more repositories in a single
dashboard.
2. Implemented new methods in the reports, benchmarks, graphs, and
summary typescript files to handle new logic for comparing repos.
3. Refactored existing code by replacing commonly used strings with
constants, and added common functions in the llmUtils file.
4. In the table section of the dashboard, added a new column
"sourceRepository", which helps in identifying the repo name when
results are rendered together for multiple repos.
5. In the Graphs, added a new prefix in the fields (eg: `vllm`) to help
distinguish and compare the results better.

**TODOs (Future scope)**:
1. Add a `QPS` dropdown filter to allow the user to compare two
benchmarks with same qps value easily. Currently, it can be done by
selecting the options in the graph, but is a nice-to-have feature.

**Testing**
Verified through local development that the dashboard is correctly
rendering the results for both the repositories, including the table as
well.
Dashboard
[link](https://torchci-60fwxh8nt-fbopensource.vercel.app/benchmark/llms?repos=vllm-project%2Fvllm%2Csgl-project%2Fsglang)

`SGLang vs vLLM Comparison Dashboard`
<img width="1709" height="948" alt="Screenshot 2025-09-08 at 2 25 46 PM"
src="https://github.com/user-attachments/assets/83f9f5ce-fef9-4e38-8cf8-8014a3dbd3d9"
/>

<img width="1709" height="948" alt="Screenshot 2025-09-08 at 1 54 12 PM"
src="https://github.com/user-attachments/assets/0b0a78fa-a60f-4673-9a6c-7be58e0f17a6"
/>

Author: namanlalitnyu

torchci/components/benchmark/llms/LLMsBenchmarkPage.tsx

@@ -24,6 +24,7 @@ import {
   HELION_BENCHMARK_NAME,
   REPO_TO_BENCHMARKS,
 } from "lib/benchmark/llms/common";
+import { LLMsBenchmarkMode } from "lib/benchmark/llms/types/benchmarkMode";
 import { LLMsBenchmarkProps } from "lib/benchmark/llms/types/dashboardProps";
 import { getBenchmarkDropdownFeatures } from "lib/benchmark/llms/utils/dashboardPickerUtils";
 import {
@@ -32,7 +33,7 @@ import {
 } from "lib/benchmark/llms/utils/llmUtils";
 import { LLMsDashboardPicker } from "./components/dashboardPicker/LLMsDashboardPicker";
 import { LLMsTimeRangePicker } from "./components/dashboardPicker/LLMsTimeRangePicker";
-import LLMsReport from "./components/LLMsReport";
+import LLMsReport from "./components/report/LLMsReport";
 
 export default function LLMsBenchmarkPage() {
   const router = useRouter();
@@ -44,6 +45,7 @@ export default function LLMsBenchmarkPage() {
   const initialPropsState: LLMsBenchmarkProps = {
     repoName: DEFAULT_REPO_NAME,
     benchmarkName: "",
+    mode: LLMsBenchmarkMode.General,
     modelName: DEFAULT_MODEL_NAME,
     backendName: DEFAULT_BACKEND_NAME,
     modeName: DEFAULT_MODE_NAME,
@@ -58,6 +60,7 @@ export default function LLMsBenchmarkPage() {
     rCommit: "",
     lBranch: MAIN_BRANCH,
     rBranch: MAIN_BRANCH,
+    repos: [],
   };
 
   const [props, dispatch] = useReducer(propsReducer, initialPropsState);