Fix/efp relative mode per group control from global control

[VinLau]

Fix eFP relative mode per-group control

Problem

The eFP relative color mode and log2 fold change tooltip were using data.control — the average of all group controls across the entire experiment — instead of each tissue's own group.control. This produced incorrect colours and values in any experiment where groups have different controls.

Example: Biotic Stress Botrytis cinerea eFP, AT3G24650

The 18hr group has a control mean of 6.59, the 48hr group has 10.77. The old code averaged these to ~8.68 and used that for everything.

| Tissue | Expected log2 FC | Before | After |
| Mock Treatment (Control) 18hr | 0.00 | -0.40 | 0.00 |
| Treated 18hr | -0.15 | -0.55 | -0.16 | *off by 0.01 due to rounding
| Control 48hr | 0.00 | 0.31 | 0.00 |

XML: https://bar.utoronto.ca/eplant/data/experiment/efps/BioticStressBotrytiscinerea/Arabidopsis_thaliana.xml

The control tissues in each group should always show a fold change of 0.00 and render as neutral/yellow in relative mode. They didn't.

Fix

Commit 1 — per-group control for relative mode

• EFPTooltip.tsx: tooltip log2 fold change now uses group.control (falls back to data.control, then 1)
• svg.tsx: useStyles passes group instead of data to getColor, so the relative-mode extremum calculation scales against the group's own min/max and control
• legend.tsx: removed unused EFPState import; added a comment noting the legend is a global approximation that may not exactly match per-group colours

Commit 2 — preserve absolute mode scaling

• The first commit broke absolute mode: passing group to getColor meant absolute colours normalised against group.max instead of the experiment-wide data.max, making every tissue appear fully red
• Added an optional absoluteMax parameter to getColor; absolute mode now uses absoluteMax ?? group.max
• useStyles passes data.max as the new argument
• legend.tsx unchanged — it already passes data as its group arg, so the fallback resolves correctly

What was NOT changed

• The XML parser — group-level <control> tags are parsed correctly
• The data shape in types.tsx or the loader in views/eFP/index.tsx — per-group control was already computed correctly; only downstream consumers were ignoring it, so we are just making sure it gets read
• The data.control aggregate field — it remains as a fallback

Screenshots

Relative mode — before (wrong fold changes) → after (correct)

Note

The Eplant/main.tsx change (router basename) is from <other-branch-name>, included because this branch is based on it. It will disappear from this diff once that PR merges.