fix(render,input): guard the zoom gain, clip annotation captions, no-throw dump

Three findings reachable by ordinary use:

* `Controller::axis_zoom()` divided by `Plot::ui_scale()` directly, though
  `Style::device()` documents a non-positive scale as host state that must be
  tolerated. With `ui_scale == 0` and a negative drag delta, `logical_dy` was
  `-inf`, the factor `+inf`, and the guard `!(factor > 0.0)` did not reject it:
  `zoom_x_axis`/`zoom_y_axis` computed `anchor + (bound - anchor) * inf`, and
  `set_y_range()` stored it verbatim and cleared `auto_scale` -- an axis left
  non-finite until an auto-fit. A negative scale silently inverted the zoom.
  A scale that is not positive and finite now means no zoom at all.

* `draw_annotations()` stages its caption into the frame's single text batch,
  which `flush_text()` draws *after* `clear_clip()`, so a caption staged while
  the axes clip was engaged was rasterised with the viewport clip. An
  annotation anchored outside the visible range kept a caption with no cross,
  and in a `Figure` -- whose rows tile the figure -- that caption painted into
  the neighbouring row. A caption whose anchor is outside `axes` is now skipped
  with its cross, rather than splitting the frame's one text draw call (which
  would cost a draw call per row and mis-clip the labels staged outside `axes`).

* `dump_frame_ppm()` was the one public entry point not `noexcept`, so a
  `std::bad_alloc` crossed the API boundary. Its buffers are now sized before
  anything is on disk and a failure is reported as `ErrorCode::OutOfMemory`,
  exactly as `TrueTypeFont::from_file()` reports it.

Tests: `ui_scale` 0 and -1 leave the axis range finite and unchanged with
`auto_scale` intact; an annotation anchored outside a bottom figure row's y
range leaves no caption ink in the row above, while an in-range one still
draws; a `static_assert` pins the dump's `noexcept`.

src/render/renderer.cpp

@@ -12,6 +12,7 @@
 #include <cmath>
 #include <cstdio>
 #include <cstdlib>
+#include <new>
 #include <vector>
 
 namespace plot {
@@ -1123,6 +1124,10 @@ void Renderer::draw_measures(const Plot& plot, const PlotLayout& layout) {
 
 void Renderer::draw_annotations(const Plot& plot) {
   const f32 line_h = font_backend_.height();
+  // The clip every other overlay is drawn under, read here because the caption is staged
+  // into the frame's single text batch -- flushed after `clear_clip()` -- so the scissor that
+  // keeps the rest of this pass inside the axes does not reach it.
+  const Rect& axes = plot.layout().axes;
   for (u32 i = 0; i < plot.annotation_count(); ++i) {
     const Annotation& annotation = plot.annotation(i);
     if (!annotation.visible) continue;
@@ -1134,7 +1139,14 @@ void Renderer::draw_annotations(const Plot& plot) {
     lines_.add_segment({px - style_.device(3.0f), py}, {px + style_.device(3.0f), py});
     lines_.add_segment({px, py - style_.device(3.0f)}, {px, py + style_.device(3.0f)});
     flush_lines(pixel_mapping(), annotation.color, style_.device(1.0f));
-    if (!annotation.text.empty()) {
+    // A caption whose anchor is outside the axes is skipped with its cross, which the clip
+    // has already removed: a caption with no cross is a label for a point the frame does not
+    // show, and -- because the text batch is drawn *after* the clip is released -- it would
+    // otherwise paint outside the axes it is measured in, into the neighbouring row of a
+    // `Figure` whose rows tile the figure. Splitting the frame's one text draw call to clip
+    // it instead would cost a draw call per row and mis-clip the tick labels, which are
+    // staged outside `axes` on purpose and share the batch after this one.
+    if (!annotation.text.empty() && axes.contains(px, py)) {
       stage_text({px + style_.device(5.0f), py - 0.5f * line_h}, annotation.text.c_str());
     }
   }
@@ -1262,7 +1274,7 @@ void Renderer::draw_crosshair_readout(const Plot& plot, const PlotLayout& layout
 
 // -------------------------------------------------------------------- dump
 
-bool Renderer::dump_frame_ppm(const char* path, Status* st) const {
+bool Renderer::dump_frame_ppm(const char* path, Status* st) const noexcept {
   if (path == nullptr || path[0] == '\0') {
     return fail(st, ErrorCode::InvalidArgument, "PPM path is empty");
   }
@@ -1275,9 +1287,20 @@ bool Renderer::dump_frame_ppm(const char* path, Status* st) const {
     return fail(st, ErrorCode::InvalidArgument, "no frame has been drawn yet");
   }
 
-  // RGBA, not RGB: 4-byte pixels have no row padding whatever the pack alignment,
-  // so a host's GL_PACK_ALIGNMENT cannot make the read overrun the buffer.
-  readback_.assign(usize(width) * height * 4u, 0u);
+  // The two buffers are sized -- and can fail to size -- before anything is on disk, so an
+  // allocation failure is reported through the `Status` rather than throwing `std::bad_alloc`
+  // across the public boundary (which is `noexcept`) and leaving a half-written file behind.
+  // The readback hole is RGBA, not RGB: 4-byte pixels have no row padding whatever the pack
+  // alignment, so a host's GL_PACK_ALIGNMENT cannot make the read overrun the buffer. The row
+  // is the PPM's own 3-byte stride, one line at a time.
+  std::vector<u8> row;
+  try {
+    readback_.assign(usize(width) * height * 4u, 0u);
+    row.assign(usize(width) * 3u, 0u);
+  } catch (const std::bad_alloc&) {
+    return fail(st, ErrorCode::OutOfMemory, "out of memory sizing the PPM readback");
+  }
+
   gl::glPixelStorei(gl::GL_PACK_ALIGNMENT, 4);
   gl::glReadPixels(static_cast<gl::GLint>(gl_viewport_.x), static_cast<gl::GLint>(gl_viewport_.y),
                    static_cast<gl::GLsizei>(width), static_cast<gl::GLsizei>(height), gl::GL_RGBA,
@@ -1288,7 +1311,6 @@ bool Renderer::dump_frame_ppm(const char* path, Status* st) const {
     return fail(st, ErrorCode::InvalidArgument, "cannot open the PPM path for writing");
   }
   std::fprintf(file, "P6\n%u %u\n255\n", width, height);
-  std::vector<u8> row(usize(width) * 3u);
   bool ok = true;
   for (u32 y = 0; y < height && ok; ++y) {
     // glReadPixels hands back rows bottom-up; a PPM's rows run top-down.