avoid quadratic group value recomputation in TokenList

[alhudz]

group_tokens and TokenList.init recompute a group's value by deep-flattening its whole subtree, so grouping an N-element identifier list is O(N^2): a 4000-column SELECT (~23 KB, under the 10000-token cap) takes ~1.8s to parse. Build the value from the direct children's already-current values and append only the newly added tokens on extend. Linear time, byte-identical parse output, full test suite passes.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 97.11%. Comparing base (7334ac9) to head (a116cbe).
⚠️ Report is 100 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #848      +/-   ##
==========================================
+ Coverage   97.04%   97.11%   +0.06%     
==========================================
  Files          20       31      +11     
  Lines        1558     3642    +2084     
  Branches        0      327     +327     
==========================================
+ Hits         1512     3537    +2025     
- Misses         46       63      +17     
- Partials        0       42      +42     

☔ 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.

[andialbrecht]

@alhudz thanks for the patch!