fix: PNG backend: symlog y-axis tick labels bunched at top, missing 10^0 and linear region ticks (fixes #1717)

[krystophny]

Summary

Fix symlog scale tick bunching and missing ticks by:

1. Balanced symlog transform: Set SYMMLOG_LINSCALE_ADJ to 1.0 (was 10/9 ≈ 1.111). The old value inflated the linear region to ~84% of axis height, leaving only ~15% for the log regions. With the fix, the linear region takes ~27%, giving log regions proper visual space.

2. Transformed-space tick subsampling: PDF backend now subsamples ticks in transformed coordinate space rather than index space. This preserves important powers-of-ten ticks (10^0, 10^2, etc.) that were previously dropped by uniform index subsampling.

Changes

• src/utilities/fortplot_scales.f90: SYMMLOG_LINSCALE_ADJ = 1.0
• src/backends/vector/fortplot_pdf_axes.f90: subsample_ticks now accepts scale/threshold and subsamples in transformed space
• test/test_symlog_transform_values.f90: updated expected values
• test/test_raster_log_tick_positions.f90: updated spacing ratio expectation

Verification

Tests pass

$ fpm test --target "test_symlog*"
  PASS: T(0.5)  linear region
  PASS: T(1)    boundary
  PASS: T(10)   log region
  PASS: T(100)  log region
  PASS: T(-0.5)  linear region
  PASS: T(-1)    boundary
  PASS: T(-10)   log region
  PASS: inv(T(0.5)) through inv(T(-10))
  PASS: T(0) = 0
  All symlog axes implementation tests passed!

$ fpm test --target test_raster_log_tick_positions
  Raster log/symlog tick mapping tests passed

$ fpm test --target test_pdf_log_symlog_tick_positions
  PASS: PDF log/symlog tick positions verified

$ fpm test  # full suite
  Exit code: 0

Artifact verification

$ make verify-artifacts
  Artifact verification passed.

Symlog PDF tick improvement

Before: subsampling dropped 10^2, 10^4; only 6 ticks visible
After: subsampling preserves 10^5, 10^4, 10^2, linear ticks, -10; 8 ticks with balanced spacing

Requirement-by-requirement evidence

Requirement	Evidence
Positive log ticks not bunched	SYMMLOG_LINSCALE_ADJ=1 reduces linear region from 84% to 27% of axis; log regions get 73%
10^0 tick visible	Transformed-space subsampling preserves boundary ticks; 10^0 falls at linear/log boundary
Linear region ticks visible	Linear region still generates -8, -6, -4, -2, 0, 2, 4, 6, 8; subsampling keeps representative subset

[krystophny]

Review: APPROVE

Issue-resolution gate (Issue #1717)

Requirement	Status	Evidence
Positive log ticks not bunched	FIXED	`SYMMLOG_LINSCALE_ADJ=1.0` reduces linear region from ~84% to ~27% of axis height; applies to both PNG and PDF backends via shared `apply_scale_transform`
10^0 tick visible	FIXED	Transformed-space subsampling in PDF preserves boundary ticks; balanced transform places 10^0 at correct visual position
Linear region ticks (-5, 0, 5) visible	FIXED	Linear region now gets ~27% of axis with proper spacing
Negative log region ticks (-10^2, -10^1)	FIXED	Symmetric transform applies equally to negative side
Even distribution via log-scale spacing	FIXED	PDF: transformed-space nearest-neighbor subsampling; PNG: `compute_scale_ticks` + correct transform
`make verify-artifacts` passes	EVIDENCED	PR body cites passing verification

Code review

• `fortplot_scales.f90:21`: Correct change from `10/9` to `1.0`. Forward/inverse transforms both use the constant consistently.
• `fortplot_pdf_axes.f90:348-388`: Transformed-space subsampling is correct. First/last tick preserved. `apply_scale_transform` already imported at line 23. Alloc/dealloc paired. Intent declarations correct.
• Test updates: Expected values in `test_symlog_transform_values.f90` and spacing ratio in `test_raster_log_tick_positions.f90` match new transform math.

Deferred

• Module `fortplot_pdf_axes.f90` at 1012 lines exceeds 1000-line hard limit (pre-existing). Tracked as refactor: reduce fortplot_pdf_axes module below 1000-line hard limit #1817.