fix: preserve Zenodo archive extraction layout

[lfvasconcelos]

Summary

• extract the Zenodo archive relative to the working directory, matching the former unzip behavior
• prevent duplication of the archive's existing data/010_eurostat_tables/ prefix
• add a cross-platform regression test that builds a representative ZIP fixture and checks the extracted path

Context

PR #12 identified that the system unzip command was not portable to Windows. The alternative crossplattform-zip branch fixed the portability issue with p7zip_jll, and commit ebfb7f4 was subsequently tested successfully on Windows.

master currently contains the earlier commit 704c5db, which passes -odata/010_eurostat_tables to 7-Zip. Because the Zenodo archive already contains that directory prefix, extraction produces data/010_eurostat_tables/data/010_eurostat_tables/... and downstream readers cannot find the files.

This change applies the extraction behavior from ebfb7f4 and adds a regression test for the actual archive layout.

Validation

• Windows 11, Julia 1.11.5
• CalibrateBeforeIT package load and precompilation: passed
• test/test_download_function.jl: 16/16 tests passed, including the regression fixture containing data/010_eurostat_tables/payload.txt
• git diff --check upstream/master...HEAD: passed

[lfvasconcelos]

Rebased onto current master (1f302d2). New head is 237cabc; the diff is unchanged, only the base moved. @zauster, the run for it is waiting on "Approve and run workflows" whenever you have a moment.

Why the earlier work did not fully close this. The crossplattform-zip branch carried two relevant commits, but only the first one, 704c5db, reached master through merge 06aeb78. PR #18 was closed rather than merged, so the follow-up ebfb7f4 never landed.

704c5db did fix the portability problem: the system unzip call is gone and p7zip_jll works on Windows. But it also introduced extract_to = "data/010_eurostat_tables" and passes it to 7-Zip as -o$(extract_to). The Zenodo archive already carries that prefix internally, so extraction now yields:

data/010_eurostat_tables/data/010_eurostat_tables/...

Every downstream reader looks under eurostat_path = "data/010_eurostat_tables" (src/CalibrateBeforeIT.jl:86, src/import_eurostat.jl:379, 02_preprocess_raw_eurostat_tables.jl:14, among others), so on current master the files sit one level deeper than anything expects. The old unzip -o $zip_file call had no output target and extracted relative to the working directory, which is why the layout was right before.

This PR restores that behavior and adds a regression test that builds a fixture archive with the real prefix and asserts the extracted path.

Two things worth flagging for review:

• I took only the download_zenodo.jl behavior from ebfb7f4, not its Project.toml change. That commit also dropped the p7zip_jll compat entry, which 238ccb5 later pinned to "17". This PR leaves Project.toml alone.
• The extract_to positional argument and the function's return value are removed, so this is a small API change. No caller in the repo uses either; 01a_download_zenodo_tables.jl calls the function with no arguments. If you would rather not touch the signature, I can keep extract_to as an optional argument defaulting to the working directory.