fix: remove sign-guarded symlog from logarithm axis to fix negative labels for sub-1 prices

[NemeZZiZZ]

Problem

The built-in logarithm y-axis template maps values through a sign-guarded symlog:

valueToRealValue: (value) => (value < 0 ? -log10(Math.abs(value)) : log10(value))
realValueToDisplayValue: (value) => (value < 0 ? -index10(Math.abs(value)) : index10(value))

For any price in (0, 1) — e.g. an altcoin trading at 0.0000123 — log10(value) is negative, so the real-space coordinate is negative while the price is positive. The inverse functions then take that negative real value to mean "negative price" and return -index10(|real|), a huge negative number.

Verified against current main (replicating YAxis.createRangeImp + createTicksImp with the template installed), visible range [0.000008, 0.000023]:

Path	Result
displayFrom / displayTo after range gap	-138922.89 / -35200.16
Rendered axis tick labels	-130000, -120000, …, -80000
Crosshair label (CrosshairHorizontalLabelView.getText → convertFromPixel)	-69929.30 (should be ≈ 0.0000143)
Overlay point value stored on draw (OverlayView → convertFromPixel)	-64399.94 (should be 0.0000155)
Public chart.convertFromPixel() API	same garbage

Any instrument priced below 1 is affected (a [0.081, 0.119] range renders ticks -12.6 … -9.6). Prices above 1 work because log10(v) > 0 keeps the sign guard on the consistent branch.

The symlog is also not injective: valueToRealValue(10) === valueToRealValue(-0.1) === 1, so no sign-guard variant can be made invertible — the guards cannot be repaired, only removed.

Fix

Make the template a pure logarithmic axis: real = log10(value), value = index10(real), unconditionally. 4 functions + createRange, one file.

After the fix, same scenarios: tick labels 0.000009, 0.000012, …, crosshair 0.0000143, overlay round-trip exact (0.0000155 → 0.0000155). Prices above 1 are bit-for-bit unaffected (the sign guards were already taking the positive branch there).

Behavior note

Ranges containing values ≤ 0 previously rendered through the quasi-symlog (approximately correct only for purely negative ranges, wrong for mixed ones: a [-5, 100] range produced tick labels 0, 40, 80…). After the fix such ranges produce NaN coordinates — the axis renders no ticks instead of wrong ones, matching the mathematical domain of a log scale (same behavior as other charting libraries, e.g. TradingView treats non-positive values as invalid on log scale).

[liihuu]

The Y-axis supports dragging and zooming; however, if negative values ​​appear, log(-xxx) cannot be calculated, resulting in a blank Y-axis.

[NemeZZiZZ]

Hi, thanks for checking this — your case is real, let me address it directly.

1. Your case: drag the log axis below zero.

You are right: with this PR, log10(negative) = NaN, and the axis goes
blank. I checked _zoomYAxis — there is no clamp, so a user can drag
below zero. This needs handling, I agree.

2. But the old code does not handle negatives either — it draws
   wrong values.

Old mapping for a negative price p: -log10(|p|).

Example — two prices, one position:

• price 10 → real log10(10) = 1
• price -0.1 → real -log10(0.1) = 1

Same position for 10 and -0.1. Crosshair and overlays cannot tell
them apart.

3. Worse: the old code breaks normal prices below $1.
   This is the bug the PR fixes (issue [Bug] 使用对数轴坐标k线，价格y轴文字（yAxis.tickText）却显示负数 #836).

• price 0.0085 → real log10(0.0085) ≈ -2.07
• the real value is negative, but the price is positive and valid
• old code sees "negative" and flips it: -10^2.07 ≈ -117

So the label shows "-117" for a 0.0085 price. Every coin under $1
shows negative labels. That is what users reported.

Live example — ARBUSDT 15m, price ≈ 0.168, but the log axis shows
-5.5…-6.2:

4. Proposal: keep pure log + clamp the range.

Instead of keeping the broken mapping, clamp from/to to positive
values in createRange. Then:

• sub-$1 prices render correctly (this PR),
• dragging below zero pins the range at a small positive value —
  the axis never goes blank, no NaN.

If you agree with the clamp approach, I will add it to this PR.