refactor: Make [non] nullable struct fields easier to create

[scovich]

What changes are proposed in this pull request?

A lot of code (especially in tests) calls StructField::new with literal values of the nullable: bool argument. Booleans are easy to misinterpret (which value means non-null field?) -- and it's hard to read when the third arg is split to its own line in nested expressions such as:

        StructField::new(
            "fileConstantValues",
            StructType::new([StructField::new(
                "partitionValues",
                MapType::new(DataType::STRING, DataType::STRING, true),
                false,
            )]),
            true,
        ),

To improve readability and make the code less error-prone, define two new helper methods/constructors for StructField: nullable and not_null, which create struct fields having the corresponding nullability.

How was this change tested?

No new functionality, and all existing unit tests still pass.

To minimize the risk of unfaithful refactoring, the change was made in four steps:

1. Use a multi-file regexp search/replace to convert simple code such like this:

   StructField::new("a", DataType::LONG, true)

   to this:

   StructField::nullable("a", DataType::LONG)

   The exact expression used was: StructField::new(\([^()]*\), true) → StructField::nullable(\1), which ignores any constructor call containing parentheses, to avoid ambiguity.
2. Use the multi-file regexp search/replace StructField::new(\([^()]*\), false) → StructField::not_null(\1), to convert simple use not_null call sites (see above for details).
3. Use an interactive multi-file search/replace StructField::new → StructField::nullable, relying on IDE parentheses matching to identify calls that pass the literal true (first pass). As a safety measure, the resulting code is compiled; all changed call sites fail to compile because of the (now unrecognized) third arg, which can then be deleted after verifying it is the literal true.
4. Use the same two-pass process for StructField::new → StructField::not_null with literal false.

Each step is its own commit, for easier verification.

[codecov]

Codecov Report

Attention: Patch coverage is 98.76543% with 3 lines in your changes missing coverage. Please review.

Project coverage is 83.66%. Comparing base (616e9ac) to head (c5a8450).
Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
ffi/src/test_ffi.rs	0.00%	3 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #646      +/-   ##
==========================================
- Coverage   83.68%   83.66%   -0.02%     
==========================================
  Files          75       75              
  Lines       16955    16909      -46     
  Branches    16955    16909      -46     
==========================================
- Hits        14188    14147      -41     
+ Misses       2102     2099       -3     
+ Partials      665      663       -2     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[zachschuermann]

whew, (checked each line haha) LGTM!

[zachschuermann]

maybe (this is a big maybe - might not be useful) we should call out in doc comments either here or in the new constructor that these exist just as shortcuts basically?

[OussamaSaoudi]

Splitting up the commits was really clever! Made reviewing a breeze. LGTM