Fix ICC rendering intent consistency and clamp sampled curves

[JimBobSquarePants]

Prerequisites

• I have written a descriptive pull-request title
• I have verified that there are no overlapping pull-requests open
• I have verified that I am following the existing coding patterns and practice as demonstrated in the repository. These follow strict Stylecop rules 👮.
• I have provided test coverage for my change (where applicable)

Description

Fixes the ColorProfileHandling.Convert discrepancy in #3198. When the source profile requests relative colorimetric intent but the destination profile header specifies perceptual intent, conversion currently mixes the two intents and applies an incorrect perceptual PCS adjustment. Resolve the source intent once and use it for both transform selection and PCS adjustment, for scalar and span conversions. Preserve behavior is unchanged.

Clamp sampled ICC curve inputs to [0, 1] before scaling to table indices, preventing extrapolation and out-of-range lookup. Correct the casing of the two IccConverterBase filenames.

Regression coverage compares scalar and span conversion against Unicolour with an explicit shared intent, and the reporter's TIFF against a LittleCMS 2.19 reference. The TIFF comparison allows 0.0001% aggregate difference; the measured difference is 0.000083%, with no channel differing by more than one 8-bit value. Add LUT endpoint coverage for out-of-domain and non-finite inputs. Update the existing YCCK JPEG reference with the reviewed output from the corrected conversion.

Validation:

• Full Release suites on .NET 8 and .NET 10 each completed with 47,832 passed, 11 skipped, and one YCCK JPEG reference mismatch before that reference was promoted.
• The YCCK failure was reproduced in isolation on .NET 8. Its reviewed output was promoted and verified byte-for-byte; the test has not been rerun after promotion.
• The TIFF regression passes on .NET 8 Release with the normal allocator and the final 0.0001% tolerance.

Overlap check: #3023 targets main and changes the same ICC base files for nullability cleanup; it does not implement this rendering-intent fix. This PR targets release/4.1.x.