Event Deltas: Heatmap symlog scale for zero-inclusive metrics

Context

The current log-scale heatmap uses greatest(value, effectiveMin) to avoid log(0). This clamps all near-zero values into the bottom bucket, losing resolution. Similarly, values above p99 land in an overflow bucket at the top.

Proposal

Use a symlog (symmetric log) scale that smoothly transitions through zero:

• sign(x) * log1p(|x|) — handles zero and negative values naturally
• No clamping needed, so no data loss at the extremes
• Well-established in matplotlib, d3, etc.

This would eliminate the need for effectiveMin and the greatest() workaround, while preserving log-scale benefits for the bulk of the distribution.

Bundle with

• #1914 — overflow bucket indicators (related: once symlog removes clamping, overflow buckets become less of an issue, but visual indicators for quantile-clipped extremes are still valuable)

Locate the current log-scale heatmap implementation and the use of effectiveMin and greatest(value, effectiveMin). Review how zero, negative, and extreme values are currently bucketed, then determine how a symlog scale and the related #1914 overflow indicators should behave. Done means zero-inclusive metrics retain resolution without clamping and the heatmap still communicates quantile-clipped extremes.