[sensor] Replace powf with pow10_int in sensor filters

Add shared pow10_int() helper to helpers.h that computes 10^exp
using iterative multiplication/division instead of powf.

Replace powf(10, exp) calls in:
- sensor/filter.cpp: ValueListFilter and RoundFilter
- helpers.cpp: normalize_accuracy_decimals (refactored to use
  pow10_int for the general case, keeping -1/-2 fast paths)

Eliminates powf/__ieee754_powf from builds where sensor filters
are the only remaining powf call site.

esphome/components/sensor/filter.cpp

@@ -2,6 +2,7 @@
 #include <cmath>
 #include "esphome/core/application.h"
 #include "esphome/core/hal.h"
+#include "esphome/core/helpers.h"
 #include "esphome/core/log.h"
 #include "sensor.h"
 
@@ -237,7 +238,7 @@ ValueListFilter::ValueListFilter(std::initializer_list<TemplatableValue<float>>
 
 bool ValueListFilter::value_matches_any_(float sensor_value) {
   int8_t accuracy = this->parent_->get_accuracy_decimals();
-  float accuracy_mult = powf(10.0f, accuracy);
+  float accuracy_mult = pow10_int(accuracy);
   float rounded_sensor = roundf(accuracy_mult * sensor_value);
 
   for (auto &filter_value : this->values_) {
@@ -469,7 +470,7 @@ optional<float> ClampFilter::new_value(float value) {
 RoundFilter::RoundFilter(uint8_t precision) : precision_(precision) {}
 optional<float> RoundFilter::new_value(float value) {
   if (std::isfinite(value)) {
-    float accuracy_mult = powf(10.0f, this->precision_);
+    float accuracy_mult = pow10_int(this->precision_);
     return roundf(accuracy_mult * value) / accuracy_mult;
   }
   return value;

esphome/core/helpers.cpp

@@ -468,20 +468,15 @@ ParseOnOffState parse_on_off(const char *str, const char *on, const char *off) {
 
 static inline void normalize_accuracy_decimals(float &value, int8_t &accuracy_decimals) {
   if (accuracy_decimals < 0) {
-    // Clamp to -9 to keep divisor within uint32_t range (max 10^9 = 1,000,000,000)
-    int8_t dec = accuracy_decimals < -9 ? (int8_t) -9 : accuracy_decimals;
-    uint32_t divisor;
-    if (dec == -1) {
-      divisor = 10;
-    } else if (dec == -2) {
-      divisor = 100;
+    float divisor;
+    if (accuracy_decimals == -1) {
+      divisor = 10.0f;
+    } else if (accuracy_decimals == -2) {
+      divisor = 100.0f;
     } else {
-      divisor = 1000;
-      for (int8_t i = dec + 3; i < 0; i++)
-        divisor *= 10;
+      divisor = pow10_int(-accuracy_decimals);
     }
-    auto divisor_f = static_cast<float>(divisor);
-    value = roundf(value / divisor_f) * divisor_f;
+    value = roundf(value / divisor) * divisor;
     accuracy_decimals = 0;
   }
 }

esphome/core/helpers.h

@@ -439,6 +439,21 @@ template<size_t STACK_SIZE, typename T = uint8_t> class SmallBufferWithHeapFallb
 /// @name Mathematics
 ///@{
 
+/// Compute 10^exp using iterative multiplication/division.
+/// Avoids pulling in powf/__ieee754_powf (~2.3KB flash) for small integer exponents.
+/// Exact for non-negative exponents up to 10 (powers of 10 up to 10^10 are exact in float).
+inline float pow10_int(int8_t exp) {
+  float result = 1.0f;
+  if (exp >= 0) {
+    for (int8_t i = 0; i < exp; i++)
+      result *= 10.0f;
+  } else {
+    for (int8_t i = exp; i < 0; i++)
+      result /= 10.0f;
+  }
+  return result;
+}
+
 /// Remap \p value from the range (\p min, \p max) to (\p min_out, \p max_out).
 template<typename T, typename U> T remap(U value, U min, U max, T min_out, T max_out) {
   return (value - min) * (max_out - min_out) / (max - min) + min_out;