Add EmbeddedLLM supporter quote (Pin Siang Tan) (#417)

* feat(quotes): add Pin Siang Tan (EmbeddedLLM) supporter quote

* fix(quotes): use black EmbeddedLLM wordmark so theme inversion is correct

Logo was white, which rendered invisible/backwards: white in light mode and black in dark mode. Display applies grayscale + dark:invert and expects black-on-transparent base logos (matching all other org logos). Recolored to black so it shows black in light theme and inverts to white in dark theme.

* fix(quotes): correct typo in Sriram Krishnan quote ("We want to the" -> "We want the")

* fix(quotes): crop Stanford logo viewBox to remove excess whitespace

The Stanford SVG had a viewBox (-96 -33.5 832 201) with large padding around
the wordmark, so it rendered ~24% of available height and looked much smaller
than other logos. Tightened viewBox to true content bounds (640x134) with even
6-unit padding, bringing it to ~91% fill, in line with other wordmark logos.

* fix(quotes): crop AMD and Supermicro logo viewBoxes to remove whitespace

Both logos sat in a 24x24 viewBox with vertical whitespace, rendering smaller
than peers (AMD ~24% fill, Supermicro ~52%). Tightened viewBoxes to true
content bounds with even padding (AMD ~86%, Supermicro ~93% fill).

* fix(quotes): make GPU Mode logo transparent so theme inversion is correct

The PNG was fully opaque with a solid black background, so the grayscale +
dark:invert display showed a black square in light theme and a white square in
dark theme. Removed the background (alpha from luminance) and recolored the
wordmark to black-on-transparent, matching the other logos: dark in light
theme, inverts to light in dark theme.

* fix(quotes): balance logo sizes (enlarge GPU Mode; shrink AMD, SGLang, GMI)

Logos are displayed at a fixed height with object-contain, so apparent size is
driven by each file's internal padding. GPU Mode was tiny (~25% fill) -> cropped
transparent margins (~79%). AMD, SGLang, and GMI were edge-to-edge (85-96%) ->
added even padding to bring them to ~68-78%, in line with peer logos (~75%).

Author: JordanNanos