[IMP] Stop indexing module prose: drop .rst, .md and .css

[moylop260]

Why

SOURCE_GLOBS had never been justified against the graph it produces. Measured on a real
Odoo 19 instance (188 modules, 5016 files, 77935 DEFINES edges), counting how many
definitions each extension contributes:

ext	files	definitions	per file	share of the graph
.py	2755	55603	20.2	71.3%
.xml	1883	14844	7.9	19.0%
.js	890	6368	7.2	8.2%
.scss	232	912	3.9	1.2%
.sql	20	35	1.8	0.04%
.csv	86	86	1.0	0.1%
.rst	73	73	1.0	0.1%
.css	8	8	1.0	0.01%
.md	2	6	3.0	0.01%

.rst and .css build exactly one node each — the File node itself, and nothing inside
it. .md only adds the headings of a README: every Section node in that graph (4 of them)
came from its two markdown files. All three are module prose, so the graph gains a path and
no structure.

What

SOURCE_GLOBS becomes .py .xml .js .scss .csv .sql. The reasoning and the measurement
live next to the constant, so the next person changing it argues with numbers.

Why .csv stays even though it scores the same 1.0

Because structure is not the only thing an indexed file is good for: search_code greps the
contents of indexed files. Searching the graph above for access_ restricted to .csv
returns 52 hits with line numbers, e.g.

extra_addons/enterprise/ai/security/ir.model.access.csv  1-14

ir.model.access.csv is a file people really do search, so dropping it would cost a real
capability to save nothing. .rst/.md/.css are greppable too, but module prose is not
what anyone greps this graph for.

Not a memory change

Worth being explicit, because it is the obvious thing to assume: this does not make a
large scope fit. The three dropped extensions are 1.4% of the files of an instance, where
.js alone is 15%. Indexing peak tracks extracted nodes, and those come overwhelmingly from
.py. This is about keeping the graph free of prose.

Verified

• 50 tests pass, pre-commit run --all-files clean.
• New tests pin the three exclusions, the .csv exception, and that *.scss survives — it
  ends with the same three letters as *.css, so a naive removal would take it too.
• All the numbers above come from query_graph over a graph indexed inside the
  irc190_01 container, not from an estimate.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 92.86%. Comparing base (d206ab0) to head (1726a74).

Additional details and impacted files

@@            Coverage Diff             @@
##             main       #2      +/-   ##
==========================================
+ Coverage   92.73%   92.86%   +0.12%     
==========================================
  Files           6        6              
  Lines         785      799      +14     
==========================================
+ Hits          728      742      +14     
  Misses         57       57              

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.