2/2: Add enumerate() to DSLX stdlib

[joannabrozek]

The second of two PRs responsible for migrating from the built-in enumerate, which cannot be lowered to IR.

This PR includes: removing the build-in enumerate.

[mikex-oss]

Aside from obvious updates to use std:enumerate, I'm seeing the following error in a bunch of our targets.

Not sure yet how to repro this in a small standalone test.

xls/dslx/stdlib/std.x:20:13-20:19
0018: 
0019: pub fn enumerate<T: type, N: u32>(x: T[N]) -> (u32, T)[N] {
0020:     for (i, result) in 0..N {
~~~~~~~~~~~~~~~~~~^----^ TypeInferenceError: TypeInferenceError: xls/dslx/stdlib/std.x:20:13-20:19 Annotated array size is too small for explicit element count.. The body of function `enumerate` does not actually return the function's declared return type, which is `(u32, u32)[N]`
0021:         update(result, i, (i, x[i]))
0022:     }([(u32:0, zero!<T>()), ...])
Error: INVALID_ARGUMENT: TypeInferenceError: xls/dslx/stdlib/std.x:20:13-20:19 TypeInferenceError: xls/dslx/stdlib/std.x:20:13-20:19 Annotated array size is too small for explicit element count.. The body of function `enumerate` does not actually return the function's declared return type, which is `(u32, u32)[N]`
=== Source Location Trace: === 
xls/dslx/errors.cc:48
xls/dslx/type_system_v2/type_annotation_resolver.cc:325
xls/dslx/type_system_v2/inference_table_converter_impl.cc:1077
xls/dslx/type_system_v2/inference_table_converter_impl.cc:269
xls/dslx/type_system_v2/inference_table_converter_impl.cc:884
xls/dslx/type_system_v2/inference_table_converter_impl.cc:245
xls/dslx/type_system_v2/inference_table_converter_impl.cc:884
xls/dslx/type_system_v2/inference_table_converter_impl.cc:245
xls/dslx/type_system_v2/inference_table_converter_impl.cc:2007
xls/dslx/type_system_v2/inference_table_converter_impl.cc:1750
xls/dslx/type_system_v2/inference_table_converter_impl.cc:1835
xls/dslx/type_system_v2/inference_table_converter_impl.cc:737
xls/dslx/type_system_v2/inference_table_converter_impl.cc:245
xls/dslx/import_routines.cc:247
xls/dslx/type_system_v2/populate_table_visitor.cc:136
xls/dslx/type_system_v2/populate_table_visitor.cc:1939
xls/dslx/type_system_v2/typecheck_module_v2.cc:89
xls/dslx/import_routines.cc:247
xls/dslx/type_system_v2/populate_table_visitor.cc:136
xls/dslx/type_system_v2/populate_table_visitor.cc:1939
xls/dslx/type_system_v2/typecheck_module_v2.cc:89
xls/dslx/import_routines.cc:247
xls/dslx/type_system_v2/populate_table_visitor.cc:136
xls/dslx/type_system_v2/populate_table_visitor.cc:1939
xls/dslx/type_system_v2/typecheck_module_v2.cc:89
xls/dslx/parse_and_typecheck.cc:126
xls/dslx/ir_convert/ir_converter.cc:697
xls/dslx/ir_convert/ir_converter_main.cc:159

[erinzmoore]

I believe I was able to reproduce the issue by testing with an empty array, for example:

#[test]
fn emumerate_test() {
    ...

    let empty_array : u32[0] = [];
    let enumerated_empty = enumerate(empty_array);
    assert_eq(enumerated_empty, []);
}

I tried to see if I could fix this with a const if in the enumerate implementation and it also failed, which may be a separate issue:

pub fn enumerate<T: type, N: u32>(x: T[N]) -> (u32, T)[N] {
    const if N == 0 {
      (u32, T)[0]:[]
    } else {
      for (i, result) in 0..N {
          update(result, i, (i, x[i]))
      }((u32, T)[N]:[(u32:0, zero!<T>()), ...])
    }
}

[joannabrozek]

@erinzmoore, thank you for the test to reproduce the issue.

I've added more commits to this PR to provide support for empty array enumeration:

• Updated std::enumerate to use the zero macro
• Added two more test cases (enumeration of struct and empty array)
• Allowed empty range for loops (mentioned here: [enhancement] dslx should support empty for loops #3965)
• Allowed indexing into zero-sized arrays for constexpr

Please let me know if this solved the issue on your side as well.