Expose OMML equation content (m:oMath) - equations are silently dropped from Paragraph.text #95

Word stores equations as OMML — m:oMath inline in a paragraph, or m:oMathPara as a block-level sibling of w:p. Neither is reachable from the API, and the m namespace (http://schemas.openxmlformats.org/officeDocument/2006/math) is not even in oxml/ns.py's nsmap.

This is silent data loss rather than a missing convenience. Equation text lives in m:t inside m:r, not in w:t inside w:r, so:

• Paragraph.text drops every equation in the paragraph with no indication one was there
• Paragraph.runs and iter_inner_content() skip them
• replace_text() and isolate_run() offsets are computed over text that does not match what the document shows

A document like The result is <equation> for all n reads back as The result is for all n.

(Font.oMath already exists but is unrelated — that is w:rPr/w:oMath, the flag marking a run as part of an equation, not the equation content.)

Where the prior art is

VanOord/vo-python-docx, branches equation_support and update_package, adds docx/oxml/math.py with CT_OMath and CT_OMathPara, docx/text/math.py with an Equation proxy exposing .raw_xml, and Paragraph.equations returning [Equation(eq, self) for eq in self._p.oMathPara_lst].

That is the minimum viable version — find the equations, hand back their XML. It is a reasonable first target; a full OMML object model is a much larger job and probably not worth it.

Shape

• m in nsmap, and the schema is already vendored at ref/xsd/shared-math.xsd
• CT_OMath / CT_OMathPara registered in oxml/__init__.py
• Math proxy in a new docx/math.py with at least .xml and .text (the concatenation of the m:t descendants)
• Paragraph.math / Document.math, and inclusion in Paragraph.iter_inner_content()

Decide explicitly

Whether Paragraph.text should include equation text. Including it is more truthful about the document's content and keeps replace_text() offsets honest, but it is a behaviour change for anyone parsing documents with equations. Given 2.0.0 already redefined Paragraph.text for tracked changes and w:fldSimple, doing it here in the same release is the cheaper moment. Whatever is chosen needs a HISTORY.rst entry.