Add configurable log path replacement control

[wolfv]

Summary

This PR introduces a new LogPathReplacements enum that provides fine-grained control over which filesystem paths are replaced with variable names (like $PREFIX, $SRC_DIR) in script output logs. This allows users to see concrete filesystem paths when needed for debugging.

Key Changes

• New LogPathReplacements enum in rattler_build_script/src/execution.rs:

  • All: Replace PREFIX, BUILD_PREFIX, and SRC_DIR (default for build scripts)
  • ExceptSrcDir: Replace PREFIX and BUILD_PREFIX only (default for test scripts)
  • None: Disable all path replacements (for debugging)

• Updated ExecutionArgs to include log_path_replacements field and modified the replacements() method to respect the configured replacement level

• Added --no-log-path-replacement CLI flag in CommonOpts to allow users to disable path replacements from the command line

• Updated TestConfiguration with:

  • New no_log_path_replacement field
  • Helper method log_path_replacements() that returns ExceptSrcDir by default (since test working directories are temporary) or None if disabled

• Updated BuildConfiguration with no_log_path_replacement field to control path replacement behavior during builds

• Propagated the setting through all script execution paths:

  • Build scripts use LogPathReplacements::All by default
  • Test scripts use LogPathReplacements::ExceptSrcDir by default
  • Both respect the no_log_path_replacement flag to disable all replacements

• Updated Python bindings to pass the flag through the CLI API

Implementation Details

The replacements() method now conditionally builds the replacement map based on the LogPathReplacements setting, allowing different levels of path obfuscation. Test scripts specifically skip SRC_DIR replacement by default since the working directory during tests is a temporary directory, not the actual source directory.

https://claude.ai/code/session_013oRb3swxDZ3guXSxe7tSgc

[wolfv]

I don't think this addresses the root cause of the issue properly.

I noticed (when looking at Jinja variabesl / SP_DIR) that there are a few gaps that we need to look at.

First of all, I think Jinja should work in tests, too. We should make the script variables and variant values available at test time as well.

The $SRC_DIR_run_env is only a problem if it's a noarch: python (or maybe noarch: generic) package.

tests:
  - script:
    - echo "PREFIX = $PREFIX"
    - echo "JINJA_PREFIX = ${{ PREFIX }}"
    - echo "SP DIR = $SP_DIR"
    - echo "And from Jinja = ${{ SP_DIR }}"

results in:

 │ Installing test environment
 │ ✔ Successfully updated the test environment
 │ Testing commands:
 │ PREFIX = $SRC_DIR_run_env
 │ JINJA_PREFIX = 
 │ SP DIR = $SRC_DIR_run_env/lib/python3.1/site-packages
 │ And from Jinja =