Excel extraction should produce per-sheet .csv.txt files with frontmatter

[sbaker]

Summary

When an Excel file (.xlsx) is extracted during compilation/packaging, the current implementation produces raw text output. Instead, each sheet should be extracted as a separate .csv.txt file with YAML frontmatter metadata.

Expected Behavior

Given an Excel file with sheets "Revenue" and "Expenses":

Output: revenue.csv.txt

---
source: financials.xlsx
sheet: Revenue
rows: 150
columns: 8
extracted_at: 2025-03-11T00:00:00Z
---
col1,col2,col3,...
data,data,data,...

Output: expenses.csv.txt

---
source: financials.xlsx
sheet: Expenses
rows: 75
columns: 6
extracted_at: 2025-03-11T00:00:00Z
---
col1,col2,col3,...
data,data,data,...

Current Behavior

Excel files are extracted as a single blob of text without sheet separation or frontmatter.

Context

• This applies to the registry-bound output — by the time content hits the registry it should already be in .csv.txt format
• Each sheet becomes its own file, broken out as individual code files with frontmatter
• Frontmatter should include source file, sheet name, and basic stats
• The xlsx dependency (0.18.5) is only used for trusted text extraction during compile/package, not user-uploaded content

Affected Components

• Python CLI: compiler.py (binary asset extraction stage)
• Go CLI: compiler.go
• Node.js CLI: compilation pipeline