feat(compose): read stdin content from -f -

[jahirvidrio]

What problem does this solve?

cat docker-compose.yml | mocker compose config -f - fails with Error: no configuration file provided: not found: - and exit 1. Docker compose accepts stdin via -f - as documented in its CLI reference, so any user migrating a shell script or a CI/CD pipe hits a Docker-parity gap on their first invocation against mocker.

The same trap exists for heredocs (mocker compose -f - config <<'EOF' ... EOF), for piping templated output (gomplate ... | mocker compose -f - up), and for mixing a file with an override supplied via stdin (-f base.yml -f - to merge). Today every one of these paths lands on the same not found: - error because loadCompose() calls ComposeFile.load(from: "-", ...) which has no stdin-reading branch.

What changed

• Add a public ComposeFile.load(content:projectDir:) overload that takes an in-memory YAML string. The .env discovery and variable substitution pipeline are factored into a private parseAndSubstitute(content:projectDir:) helper that the existing load(from:projectDir:) now delegates to (no behavior change to file loading).
• Wire the stdin branch in loadCompose(): when an entry in -f is "-", read with FileHandle.standardInput.readToEnd() (mirroring Login.swift:30's established pattern) and route the bytes through the new overload. Absolute paths and the default-file path pass through unchanged.
• Empty or whitespace-only stdin surfaces as MockerError.composeParseError("compose file content is empty") — the existing case, not a new enum variant. Empty stdin and empty file are the same failure mode (input that cannot be parsed), so both paths share the error category; the detail message tells the user which one fired.
• Multi-stdin (-f - -f - in the same invocation) is undefined: the second read returns empty bytes and triggers the empty-stdin error. Use each -f - once per invocation, which is what the upstream compose reference documents.