Load workspace config when indexing tracked repos

[alex4o]

Summary

• load each tracked repo's .gortex.yaml before IndexAll creates its per-repo indexer
• make server --track honor repo-local excludes and workspace config from the tracked destination repo
• add regression coverage for single-repo and multi-repo IndexAll respecting workspace excludes

Test plan

• go test ./cmd/gortex ./internal/config ./internal/indexer ./internal/excludes
• go build -o gortex ./cmd/gortex

[zzet]

Hey @alex4o!

Thank you for catching this nasty bug and contribution.

I followed further for the change and found MultiIndexer.IndexRepo at

gortex/internal/indexer/multi.go

Line 359 in a296103

cfg := mi.configMgr.GetRepoConfig(repoPrefix)

has the same bug - calls GetRepoConfig without LoadWorkspaceConfig.
It's reached from the MCP index_repository tool

gortex/internal/mcp/tools_core.go

Line 552 in a296103

result, err := s.multiIndexer.IndexRepo(prefix)

On a fresh-start daemon where immediately calls index_repository against a repo with a .gortex.yaml, you'd still get the bug. I believe it make sense to patch it the same way for completeness.
Also, tests only cover the exclude symptom and they don't lock down the workspace-slug propagation, which is the other (and arguably more impactful) thing your fix unblocks. Adding a workspace-slug assertion would harden against regressions.

Can you please take a look to those cases? If you don't have time for it - I can accept the PR and apply those patches in the follow up commit later.