Cardano Updates

Freeze the current (unhelpful) error output for three forms of invalid
UPLC identifier:
- `foo-bar`                — hyphen followed by non-digits
- `foo-123-456`            — double `-NNN` suffix
- `pubKeyHash-305478r71`   — hyphen + digits + more letters (the shape
  Scalus 0.16.0's `toUplcOptimized` emits, from issue #7742)

All three cases produce misleading diagnostics today — notably the
Scalus case reports the error 8+ characters past the offending name.
Capturing the status quo as goldens so that a follow-up improvement to
name-parser diagnostics shows up as an explicit golden-file diff.

When the unquoted-identifier parser finishes, require that the next char
is a real word-boundary (not another identifier char and not another
'-'). Otherwise the caller wrote something like `pubKeyHash-305478r71`,
`foo-bar` or `foo-123-456`: the '-NNN' we just consumed as the numeric
unique-suffix is not actually terminal, and the prefix interpretation
would silently mis-parse. Consume the remainder of the extended
identifier so the diagnostic can cite the full bad text, then raise a
new `InvalidIdentifier` custom parser error with a caret on the start
of the identifier and an actionable hint to quote it with backticks.

For the original Scalus 0.16.0 HTLC reproducer this changes the error
from `htlc.uplc:448:39: unexpected '(' expecting ')'` (on a lambda 8+
chars past the real site) to `htlc.uplc:447:41: Invalid identifier
'pubKeyHash-305478r71'` — on the offending name itself.

The three negative goldens added in the previous commit are updated to
the new message; all 3886 tests across plutus-core/untyped-plutus-core/
plutus-ir pass unchanged.

`makeSizedTextStrings` and `makeSizedUtf8ByteStrings` had the same bug as
`makeSizedByteStrings`: they fmap'd a single seed across all sizes,
producing correlated string and decodeUtf8 benchmark inputs.

Extract the seed stream into a named top-level binding `splitSeeds` and
reuse it across all three sized-list helpers.

When the unquoted-identifier parser finishes, require that the next char
is a real word-boundary (not another identifier char and not another
'-'). Otherwise the caller wrote something like `pubKeyHash-305478r71`,
`foo-bar` or `foo-123-456`: the '-NNN' we just consumed as the numeric
unique-suffix is not actually terminal, and the prefix interpretation
would silently mis-parse. Consume the remainder of the extended
identifier so the diagnostic can cite the full bad text, then raise a
new `InvalidIdentifier` custom parser error with a caret on the start
of the identifier and an actionable hint to quote it with backticks.

For the original Scalus 0.16.0 HTLC reproducer this changes the error
from `htlc.uplc:448:39: unexpected '(' expecting ')'` (on a lambda 8+
chars past the real site) to `htlc.uplc:447:41: Invalid identifier
'pubKeyHash-305478r71'` — on the offending name itself.

The three negative goldens added in the previous commit are updated to
the new message; all 3886 tests across plutus-core/untyped-plutus-core/
plutus-ir pass unchanged.

Freeze the current (unhelpful) error output for three forms of invalid
UPLC identifier:
- `foo-bar`                — hyphen followed by non-digits
- `foo-123-456`            — double `-NNN` suffix
- `pubKeyHash-305478r71`   — hyphen + digits + more letters (the shape
  Scalus 0.16.0's `toUplcOptimized` emits, from issue #7742)

All three cases produce misleading diagnostics today — notably the
Scalus case reports the error 8+ characters past the offending name.
Capturing the status quo as goldens so that a follow-up improvement to
name-parser diagnostics shows up as an explicit golden-file diff.

Nix's default `.#` flake ref treats the flake as a Git-tracked input; in
bare-repo + worktree setups this produces "Path 'flake.nix' is not
tracked" errors even though the file is tracked. The `path:` prefix
sidesteps Git and reads the directory directly, with identical output
in standard clones. Encountered during Plutus 1.62.0.0 release, step 6
(`publish-gh-release`).

* Move CSE to the unimplemented-in-the-certifier category

* Remove dead code

* Fix order of constructors in Agda <-> Haskell FFI