Fix tdigest

[LiaCastaneda]

Which issue does this PR close?

• Closes #.

Rationale for this change

SELECT 
  approx_percentile(x, 0.5) AS p50,
  approx_percentile(x, 0.90) AS p90,
  approx_percentile(x, 0.99) AS p99
FROM (VALUES (1), (2), (3), (4), (5), (10), (20), (50), (100), (1000)) AS t(x)

returns 6, 550, 100, where p90>p99, which look like wrong results.

What changes are included in this PR?

Adds a boundary check in estimate_quantile() when rank >= count - 1.0, returns max directly instead of interpolating. Without this, the interpolation could produce values larger than higher quantiles, causing p90 > p99 on sparse datasets.

Are these changes tested?

Added test_sparse_dataset_quantile_ordering

Are there any user-facing changes?