chore: slim analytics package to the static-site core

[NoopDog]

Part 3 of retiring the legacy analytics report formats — do this only after #4910 and #4911 are merged (they remove the in-repo consumers of the code deleted here). Completing the dependency-burden rationale from #4910: after this ticket, the package's install_requires no longer pulls gspread, gspread-formatting, or matplotlib, and consuming repos (anvil-portal, ncpi-dataset-catalog) can drop those pins from their dependabot-watched requirements.

Current structure (verified)

• sheets_api.py (493 lines) — pure gspread/Drive/Sheets formatting; only legacy consumers.
• sheets_elements.py (606 lines) — mostly GA4 element functions the static site needs (get_page_views_change, get_page_views_over_time_df, get_outbound_links_change, get_index_filter_selected_change, get_index_table_download_df, get_bounds_for_month_and_prev, ADDITIONAL_DATA_BEHAVIOR), plus Sheets-only *_FORMATTING dicts importing COLUMN_FORMAT_OPTIONS from sheets_api.
• charts.py (454 lines) — core GA fetch helpers (get_data_df, get_df_over_time) needed by _sheets_utils, mixed with matplotlib/IPython notebook plotting.
• api.py, entities.py, _sheets_utils.py, static_site/ — the keep-side core.

Steps

1. Extract get_data_df / get_df_over_time (and any pure-fetch helpers they need) from charts.py into a new module (e.g. ga_data.py); update _sheets_utils.py imports.
2. Delete the remainder of charts.py (matplotlib/IPython plotting), and delete sheets_api.py.
3. Strip the *_FORMATTING dicts and the from .sheets_api import COLUMN_FORMAT_OPTIONS import from sheets_elements.py; rename it (e.g. elements.py) and update static_site/fetch.py (from .. import sheets_elements as elements → the new name). Optionally rename _sheets_utils.py similarly.\n4. setup.py: remove gspread, gspread-formatting, matplotlib from install_requires; bump the version (breaking change for any straggler consumer of the deleted modules).
4. Regenerate analytics/requirements.txt without the removed packages.

Verification

• python -c 'from analytics.static_site import generate_site' succeeds in a fresh venv without gspread/matplotlib installed.
• Run one full generate_static_site.py (anvil-explorer) — output data unchanged.

Downstream

After merge, anvilproject/anvil-portal (which installs this package from main) and NIH-NCPI/ncpi-dataset-catalog can finish slimming their requirements — tracked in their per-repo retirement tickets.