Use face-scaled accrued interest in bond cash values

[kadyrbekovhamit-cyber]

Description

Use consistently scaled accrued interest in four bond cash-value paths:

• Bond.principal()
• BondFRN.principal()
• InflationBond.inflation_principal()
• Bond.clean_price_from_survival_curve()

The three principal methods scale dirty price by face / par, but the released implementation subtracts accrued interest calculated for face 1.0. The unscaled subtraction violates the basic identity

principal(face) = clean_price_per_100 * face / 100

The released BondZero.principal() already follows this pattern by calculating accrued interest with the supplied face.

BondFRN.principal() and BondFRN.clean_price_from_dm() also relied on prior calls to initialize their previous/next coupon dates. Both now calculate accrued interest before dirty price, so direct calls on a newly constructed bond work without hidden state.

Separately, Bond.clean_price_from_survival_curve() returns a price per 100 of par but subtracts accrued interest for face 1.0. In the synthetic released-version reproduction it returns 103.9834175889; the dirty-minus-accrued-per-100 identity is 103.0152960420, a difference of 0.9681215470.

Released-version reproductions

Using synthetic inputs with FinancePy 1.1.2:

Bond.principal(), face 1,000,000
actual:   1,027,043.3256
expected: 1,017,264.3298
difference: 9,778.9957

BondFRN.principal(), face 1,000,000
actual:   970,266.2031
expected: 967,929.9898
difference: 2,336.2132

InflationBond.inflation_principal(), face 1,000,000
actual:   1,167,344.6587
expected: 1,167,116.0212
difference: 228.6375

These are method-level synthetic reproductions only; no production system or customer data was tested.

Correction

Each principal method now recalculates accrued interest for the supplied face amount before subtraction. The survival-curve clean-price path recalculates it for self.par. This also leaves the object's accrued-interest state consistent with the cash value being returned.

Testing

• Added regressions at face 1, 100 and 1,000,000.
• Added a fresh-object direct-call regression for BondFRN.principal().
• The same regression calls BondFRN.clean_price_from_dm() on a separate fresh object before any other method.
• Updated the existing face-100 expectations to the clean-value identities.
• Focused bond tests: 17 passed.
• Full unit suite: 962 passed, 4 pre-existing numerical-fit warnings.
• Mutation checks: restoring each old unscaled subtraction makes its new regression fail; restoring the corrections returns the focused tests to green.