Implement Aggregation/Reduce feature

[vigith]

Ability to implement stream aggregation or reduction by providing user-defined functions.

spec:
  vertices:
    - name: reduce-by-key
      udf:
        container:
          image:sum-by-key:latest
        groupby:
           window:
             fixed:
               length: 1m # 1 minute
           # if true, then it is a keyed stream, else it is non-keyed.
           keyed: true # can be false too

• feat: reducer for stream aggregation without fault tolerance #208
• reduce end to end test using in-memory watermark #233
• windowing strategy as an interface #234
• Implement sliding window #339
• move watermark at millisecond granularity #201
• Support Reduce Vertex #246
• integrate reduce vertex with RL and PnF #289
• Add shuffle in WhereToStep #291
• feat: add pvc support for reduce vertex PBQ #292
• Unit test core code paths in reduce #373
• Too many keys in Offset Timeline bucket causing slowness in watch and publisher startup #301
• Integrate with WAL
  • Implement WAL with one segment that support read/write paths #343
  • Implement WAL, make PBQ (persistent buffered queue) fault tolerant #293
• Export prometheus metrics for reduce processing #313
• End to End test for Reduce #338
• Unary gRPC message size limit 4MB #237
• SDK Support for reduce
  • Golang
  • Python
  • Java
• FMEA
• Documentation for reduce #403
  • user guide
  • example pipelines
• Blog for reduce