Parquet file content is different if ~/.cargo is in a git checkout

[carols10cents]

Describe the bug

I check my home directory into git. My home directory contains .cargo, my CARGO_HOME directory. When I write a Parquet file, its FileMetaData contains:

created_by: Some(
    "parquet-rs version 5.0.0 (build 3ef76a677716df403a13964a58351abe37c1754d)",
),

That SHA is of a commit in my home directory, not in Parquet, and not in the project using Parquet.

I have a test in the project that verifies the size of the parquet file data, and the test was failing for me because the content was 49 bytes too much, the exact size of the extra content above. I verified that in CI, the test passes, and the FileMetaData under test contains:

created_by: Some(
    "parquet-rs version 5.0.0",
),

To Reproduce

• Check your home directory into git, or alternately set CARGO_HOME to a directory in a git repository.
• Generate a parquet file and check the metadata.
• Observe the created_by contains a hash from the git directory CARGO_HOME is in.

I'm not sure if it's going to be possible to create a failing test for this given the environmental aspect... the current test only checks that the created_at value is the value of the PARQUET_CREATED_BY environment variable but the problem is what gets in the PARQUET_CREATED_BY environment variable in the first place.

Expected behavior

I expected to get the exact same Parquet file content whether my home directory is checked into Git or not 🤣

Additional context

The PARQUET_CREATED_BY environment variable is set in the build script if git rev-parse HEAD returns a value. Considering this is only getting set if you have a non-standard setup like I do, I think this should just be removed entirely. I'm going to prepare a PR for discussion with this solution :)