feat: support fenl types in Collection

[jordanrfrazier]

FenlType::Collection currently only supports type variables, meaning we can't do map<i32, V> type signatures. We should refactor the Collection type to accept FenlTypes instead of just type vars, then clean up the parsing and inference code around that.

Tasks:

• Support map<k,v> in type inference
• Support DataType::Map in merge (spread operation)
  • Unlatched (feat: add unlatched map spread and largeutf8 spreads #529)
  • Latched
• Support DataType::Map in first/last aggregations
  • non-windowed feat: support first and last non-windowed aggregations over DataType::Map #540
  • since window
  • sliding window
• Support list<e> in type inference feat: index for DataType::List #562
• Support DataType::List in merge (spread operation)
• Support DataType::List in aggregations
  • (non-windowed) First/Last feat: add first/last list non-windowed aggregation support #592
• List construction using collect function
  • Doc: https://docs.google.com/document/d/1P6sogDYfq31n3LXe3WViDcHWGBfHRxNZwd1znjmagQk/edit
  • Record Types feat: support DataType::Struct in collect function #602
  • Primitive types feat: collect to list (non-windowed) (primitive/strings/booleans) #569
  • Complex/Nested types
  • Non-windowed feat: collect to list (non-windowed) (primitive/strings/booleans) #569
  • Since Window feat: support since windows in collect aggregations #583
  • Sliding Windows
• Map construction
  • TODO: Doc
  • Primitive Types
  • Complex/Nested Types
• Brackets for access: map[key] list[index]
• Add get by key function for maps (get(key, map)) (feat: support more types for get evaluator for maps #532)
  • Support boolean/string/primitives
  • Complex/Nested types
• Add get by index for lists feat: index for DataType::List #562
• Verify all normal functions interact with collection types as expected
• Support LargetUtf8
  • Not done in multiple aggregations yet
• Support Map Equality in Arrow eq_dyn kernel: https://docs.rs/arrow-ord/44.0.0/src/arrow_ord/comparison.rs
  • Easily done if the map is ordered, but trickier to do (efficiently) if not: [{f1: _, f2: _}] must still equal [{f2: _, f1: _}].
• Improve map evaluators -- use GetIndex and take kernel to reduce duplication of evaluators/improve iteration
• Test on other sample datasets
• Using Dictionary types for efficient List (and Map?) aggregation
• Other functions
  • keys()
  • values()
  • entries()
  • getEntryFromKey/Value()
  • list_len() feat: add list_len function #609
  • reverse()
  • column<T: Struct, N: any>(input: list<T>, field: string) -> list<N>
    • takes a list of columns and gets the value for the field for each struct in each list, then collects them to a list<N>
  • Union Lists together
    • Having lists from multiple expressions and wanting to union them together to operate on a single list.

[jordanrfrazier]

Map and list types are not displayed correctly in the schema (see readings)