fix: include all mapped columns in stat transform GROUP BY

[cpsievert]

Summary

Fixes #239.

Integer columns mapped to non-positional aesthetics like fill or color were being dropped by stat transforms (e.g., bar's COUNT). This happened because only String/Boolean/Categorical columns were considered "discrete" and added to partition_by — integer columns were skipped, causing them to be excluded from GROUP BY.

Before: DRAW bar MAPPING sex AS x, survived AS fill (where survived is integer 0/1) → validation error because fill column was dropped by stat transform.

After: All non-positional, non-stat-consumed mapped columns are added to partition_by regardless of data type, matching ggplot2's grouping behavior.

Changes

• src/execute/mod.rs: Renamed add_discrete_columns_to_partition_by → add_mapped_columns_to_partition_by and removed the discreteness filter. Simplified the function signature (removed unused scales and aesthetic_ctx parameters).
• src/execute/layer.rs, src/execute/position.rs: Updated comments referencing the renamed function.
• New test: test_bar_with_integer_fill_column verifies integer fill columns survive bar stat transforms.

Test plan

• New test test_bar_with_integer_fill_column passes
• All 1221 existing tests pass
• No compiler warnings

🤖 Generated with Claude Code

[cpsievert]

@thomasp85 Take this with a heavy grain of salt. I haven't investigated this change at all yet -- more so just proposing it to promote awareness and get a conversation going.

[thomasp85]

I'm very much against this. If you wish to use an integer column then make it discrete (which will make it part of the grouping). You are already saying that you meant for the column to be discrete but it happened to be numeric - the fix is to explicitly make it discrete

DRAW bar 
  MAPPING sex AS x, 
  survived AS fill
SCALE fill VIA bool
-- or
SCALE DISCRETE fill
-- or
SCALE ORDINAL fill

[cpsievert]

Ahh, right, I'm happy to work around this via prompting. This might a good case to keep in mind for #85?