* feat(benchmarks): add baseline performance metrics documentation

Add comprehensive BASELINE.md template for network metrics:
- Connection establishment performance (<100μs localhost target)
- Session management and pooling
- Data transfer throughput (>1GB/s localhost target)
- Round-trip latency (<50μs for 0-byte ping)
- Concurrent connection scalability (10k+ target)
- Protocol overhead (TCP vs UDP)
- Event loop performance (>1M events/sec)
- Buffer management and zero-copy operations
- Template ready for data collection

Establishes Phase 0 infrastructure for network performance tracking.
Target: <5% regression, >1GB/s throughput on localhost

Part of Phase 0 completion (Task 0.2: Baseline Performance Benchmarking)

* fix(ci): add workflow file to benchmark trigger paths

Add benchmarks.yml to pull_request path filter to ensure workflow
changes trigger benchmark execution in PRs.

Previously, the workflow only triggered on changes to benchmarks/,
src/, or include/ directories in PRs, causing it to be skipped when
only the workflow file itself was modified.

This ensures consistency between push and pull_request triggers.

* fix(ci): handle documentation-only PRs in performance regression check

This commit improves the Performance Regression Check workflow to gracefully
handle PRs that only modify documentation files (like BASELINE.md).

Changes:
- Add NETWORK_BUILD_BENCHMARKS=ON flag with fallback for base/PR builds
- Check if benchmark executable exists before attempting to run it
- Add continue-on-error to benchmark run steps
- Provide informative messages for documentation-only changes

This prevents the workflow from being marked as 'skipped' when benchmark
executables don't exist, making the CI status clearer.

* fix(ci): prevent duplicate workflow runs for PRs

Add concurrency group to prevent both push and pull_request events
from creating redundant workflow runs for the same commit.

This eliminates the confusing 'skipped' Performance Regression Check
status that appears from the push event when a PR exists.

* refactor(ci): separate performance regression into dedicated workflow

Split the Performance Regression Check from test-integration.yml into
a separate workflow file that only runs on pull_request events.

Benefits:
- Eliminates confusing 'skipped' status from push events
- Clearer separation of concerns
- Performance checks only run when needed (PRs only)
- No duplicate workflow runs for the same commit

Changes:
- Remove performance-regression job from test-integration.yml
- Create new performance-regression.yml workflow
- Configure to run only on pull_request events

* fix(ci): correct shell variable syntax in performance regression

Remove backslash escapes from shell variables. GitHub Actions YAML
only processes ${{ }} syntax, while regular shell variables should
use $ without escaping.

Fixes: syntax error near unexpected token '('