SweepFormula: fix parser handling of signed exponents, non-finite operands and empty arrays

[Copilot]

Thanks for opening a PR in MIES ✨!

• Code can only be merged if the continous integration tests pass
• Please ensure that the branch is named correctly. See
  here for the detailed explanation.

Description

Three SweepFormula parser/executor defects:

• 1e-3 gave Operand for - must be numeric (tokenizer split at the exponent sign).
• 1 - NaN / 1 - inf gave Operand for - must be numeric (non-finite operands stored as strings).
• [] (empty array) was rejected by the parser.

Changes

• Scientific notation (MIES_SweepFormula_Parser.ipf): corrected the signed-number regex exponent order (E[+-]?[0-9]+) and added a pending-exponent regex so a +/- immediately following a number ending in e/E is collected into the number rather than parsed as an operator.
• Non-finite operands (MIES_SweepFormula_Operations.ipf): text operands consisting solely of non-finite literals are converted to numeric inside the +, -, *, / operations. The recognized set (inf, -inf, NaN, -NaN, case-sensitive) mirrors the executor's SFE_ConvertNonFiniteElementsImpl so array and operation contexts stay consistent; standalone literals still return text waves.
• Empty arrays (MIES_SweepFormula_Parser.ipf): explicit empty-array construction in the array action, producing [] at top level and when nested ([1,[]], [[]]). Relaxing the shared sign-evaluation assert this requires is offset by a guard that keeps invalid () reporting a clear error instead of yielding null.
• Tests / docs: added cases in UTF_SweepFormula.ipf and documented signed exponents and non-finite operands in SweepFormula.rst.

Examples

1e-3        // 0.001
1 - NaN     // NaN
1 - inf     // -inf
[]          // empty numeric wave

[t-b]

@copilot Please fix the CI.

[Copilot]

@copilot Please fix the CI.

Fixed in 543e4a1. The compile failure was MIES_SweepFormula_Operations.ipf: error: Expected a wave reference.: the four +/-/*// impl functions redeclared their wave-reference parameters with a same-name function-call assignment (WAVE/Z data0 = SFO_ConvertNonFiniteText(data0)), which Igor parses as wave point-indexing. The converted operands are now stored in distinct locals (data0Num/data1Num).

[t-b]

@copilot Can you create a separate PR with the exponential changes?

[Copilot]

@t-b I've opened a new pull request, #2742, to work on those changes. Once the pull request is ready, I'll request review from you.

[t-b]

@copilot Please create a new MR against main which has only the changes for the "empty array" fix.

[Copilot]

@t-b I've opened a new pull request, #2749, to work on those changes. Once the pull request is ready, I'll request review from you.

[t-b]

THis will be done in separate PRs.