Add IELR report for state splits and split reasons

[ydah]

This PR adds a new --report=ielr report that makes IELR state splitting easier to inspect.

The new report shows:

• which LALR states were split by IELR
• incoming transitions for each split variant
• lookahead differences between the split states
• the reason each split happened, based on dominant actions for each token

Examples

Example output from --report=ielr for spec/fixtures/integration/ielr.y

IELR State Splits

    LALR state 4 splits into IELR states 4, 19

      Incoming transitions
        state 1 -- a --> state 4 [LALR core]
        state 2 -- a --> state 19 [IELR split]

      Lookahead differences
        a • C D E  (rule 3)
          state 4 [LALR core]: [a]
          state 19 [IELR split]: []

      Why it split
        token a
          state 4 [LALR core]: reduce using rule 9 (E)
          state 19 [IELR split]: shift and go to state 8


    LALR state 9 splits into IELR states 9, 20

      Incoming transitions
        state 4 -- C --> state 9 [LALR core]
        state 19 -- C --> state 20 [IELR split]

      Lookahead differences
        a C • D E  (rule 3)
          state 9 [LALR core]: [a]
          state 20 [IELR split]: []

      Why it split
        token a
          state 9 [LALR core]: reduce using rule 9 (E)
          state 20 [IELR split]: shift and go to state 8


    LALR state 14 splits into IELR states 14, 21

      Incoming transitions
        state 9 -- D --> state 14 [LALR core]
        state 20 -- D --> state 21 [IELR split]

      Lookahead differences
        a C D • E  (rule 3)
          state 14 [LALR core]: [a]
          state 21 [IELR split]: []

      Why it split
        token a
          state 14 [LALR core]: reduce using rule 9 (E)
          state 21 [IELR split]: shift and go to state 17

[junk0612]

The paper uses the technical term "isocore" for a set of states with the same core, and I'd like to use that term in the report as well.